from __future__ import print_function import numpy as np import sys import os import datetime import matplotlib.pyplot as plt from matplotlib.backends.backend_pdf import PdfPages from matplotlib.ticker import MultipleLocator from matplotlib.ticker import ScalarFormatter from matplotlib.ticker import NullFormatter import matplotlib.image as image import matplotlib.patches as patches import matplotlib.lines as lines import matplotlib.ticker as ticker from scipy.optimize import curve_fit from matplotlib.ticker import LogLocator, LogFormatterMathtext from Aux import * def func(x,b): return np.exp(-(x**b)) def make_figure1old (N_bunches,N_runs_bunch,N_episodes,foldername,N_bunches2,N_runs_bunch2,N_episodes2,foldername2,N_bunches3,N_runs_bunch3,N_episodes3,foldername3,testvalue,test2value): ####### Type A N = N_bunches*N_runs_bunch Nep_arrays = ep_index(N_episodes) # read average target times and number of targets episodes,avg_time,n_targets = read_AvgTargetTimes_bunches(foldername,N_bunches,Nep_arrays) # read target times and number of targets vs r_ini target_times,initial_radii,initial_directions = read_TargetTimes_bunches(foldername,N_bunches,N,Nep_arrays) target_times_ini,initial_radii_ini,initial_directions_ini = read_TargetTimes_given_policy(foldername,'InitialPolicy_old') N_ini = len(target_times_ini) ####### Type C N2 = N_bunches2*N_runs_bunch2 Nep_arrays2 = ep_index(N_episodes2) # read average target times and number of targets episodes2,avg_time2,n_targets2 = read_AvgTargetTimes_bunches(foldername2,N_bunches2,Nep_arrays2) # ~ # read target times and number of targets vs r_ini target_times2,initial_radii2,initial_directions2 = read_TargetTimes_bunches(foldername2,N_bunches2,N2,Nep_arrays2) target_times2_ini,initial_radii2_ini,initial_directions2_ini = read_TargetTimes_given_policy(foldername2,'InitialPolicy_old') N2_ini = len(target_times2_ini) ####### Type B N3 = N_bunches3*N_runs_bunch3 Nep_arrays3 = ep_index(N_episodes3) # read average target times and number of targets episodes3,avg_time3,n_targets3 = read_AvgTargetTimes_bunches(foldername3,N_bunches3,Nep_arrays3) # ~ # read target times and number of targets vs r_ini target_times3,initial_radii3,initial_directions3 = read_TargetTimes_bunches(foldername3,N_bunches3,N3,Nep_arrays3) target_times3_ini,initial_radii3_ini,initial_directions3_ini = read_TargetTimes_given_policy(foldername3,'InitialPolicy_old') N3_ini = len(target_times3_ini) for i in range(Nep_arrays3): print (ep_index(episodes[i]),episodes[i],n_targets[i]*1.0/N,n_targets2[i]*1.0/N2,n_targets3[i]*1.0/N3) for i in range(Nep_arrays3,Nep_arrays2): print (ep_index(episodes[i]),episodes[i],n_targets[i]*1.0/N,n_targets2[i]*1.0/N2,'still missing') for i in range(Nep_arrays2,Nep_arrays): print (ep_index(episodes[i]),episodes[i],n_targets[i]*1.0/N,'still missing','still missing') #################################################################### #################################################################### #################################################################### xfig,yfig=7.0,2.5 factor = xfig/yfig x1,y1 = 0.08,0.16 dx1,dy1=0.4,0.82 x2,y2 = 0.135,0.5 dx2,dy2=0.16,0.4 x3,y3 = 0.58,y1 dx3,dy3 = dx1,dy1 x4,y4 = x3+dx3,y1 dx4,dy4 = dx3,dy1 with PdfPages('fig2.pdf') as pdf: fig = plt.figure(figsize=(xfig,yfig)) plt.rc('text', usetex=True) plt.rc('text.latex', preamble = ','.join('''\usepackage{txfonts} \usepackage{lmodern}'''.split())) ######################## panel A ######################################## panel = fig.add_axes([x1, y1, dx1, dy1]) panel.tick_params(axis='both',which='both',direction='in',bottom=True,top=True,left=True,right=True) for tick in panel.xaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsize) for tick in panel.yaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsize) panel.set_xlabel(r'episodes',fontsize=axislabelfontsize) panel.set_ylabel(r'fraction of successful agents',fontsize=axislabelfontsize) # ~ panel.set_xlim(0.8,280000) panel.set_xlim(0.8,1400000) panel.set_ylim(0,1.44) panel.set_xscale('log') panel.yaxis.set_major_locator(MultipleLocator(0.1)) panel.yaxis.set_minor_locator(MultipleLocator(0.05)) panel.xaxis.set_major_locator(LogLocator(base=10.0, subs=[1.0], numticks=10)) # Major ticks at 10^x panel.xaxis.set_minor_locator(LogLocator(base=10.0, subs=np.arange(1, 10) * 0.1, numticks=100)) # Minor ticks at 1-9 panel.xaxis.set_major_formatter(LogFormatterMathtext()) # Forces 10^x notation panel.xaxis.set_minor_formatter(NullFormatter()) # plot avg line panel.plot(episodes,n_targets*1.0/N,'-',color=color1,markersize=2,linewidth=linewidth,zorder=3,label=r'type A') panel.plot(episodes3,n_targets3*1.0/N3,'-',color=color3,markersize=2,linewidth=linewidth,zorder=3,label=r'type B') panel.plot(episodes2,n_targets2*1.0/N2,'-',color=color2,markersize=2,linewidth=linewidth,zorder=3,label=r'type C') panel.hlines(y=1.0, xmin=3000., xmax=1000000, linestyle='--', linewidth=0.5, color='black') # ~ panel.hlines(y=testvalue, xmin=2000., xmax=1000000, linestyle='--', linewidth=0.5, color='brown') panel.legend(loc='upper right', bbox_to_anchor=(0.99, 0.99),ncol=1,fontsize=legendfontsize-1,handlelength=1.5,labelspacing=0.2,framealpha=1.0) panel.text(0.03,0.9,r'{\bf (a)}',fontsize=legendfontsize,transform=panel.transAxes) ######################## panel B ######################################## panel = fig.add_axes([x2, y2, dx2, dy2]) panel.tick_params(axis='both',which='both',direction='in',bottom=True,top=True,left=True,right=True) for tick in panel.xaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsizeinset) for tick in panel.yaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsizeinset) panel.set_xlabel(r'episodes',fontsize=axislabelfontsizeinset) panel.set_ylabel(r'avg. ep. time $[\tau]$',fontsize=axislabelfontsizeinset-1,labelpad=0.2) # ~ panel.set_xlim(0.8,280000) panel.set_xlim(0.8,1400000) panel.set_ylim(0.1,1.58) panel.set_xscale('log') panel.yaxis.set_major_locator(MultipleLocator(0.2)) panel.yaxis.set_minor_locator(MultipleLocator(0.1)) panel.set_xticks([1, 10, 100, 1000, 10000, 100000, 1000000]) # Specify ticks explicitly # plot avg line panel.plot(episodes,avg_time,'--',color=color1,markersize=2,linewidth=linewidth-1,zorder=3) panel.plot(episodes3,avg_time3,'--',color=color3,markersize=2,linewidth=linewidth-1,zorder=3) panel.plot(episodes2,avg_time2,'--',color=color2,markersize=2,linewidth=linewidth-1,zorder=3) ftp = (avg_time*n_targets + 1.0*(N-n_targets))/N ftp3 = (avg_time3*n_targets3 + 1.0*(N3-n_targets3))/N3 ftp2 = (avg_time2*n_targets2 + 1.0*(N2-n_targets2))/N2 panel.hlines(y=1.0, xmin=0.1, xmax=1000000, linestyle='--', linewidth=0.5, color='black') panel.plot(episodes,ftp,'-',color=color1,markersize=2,linewidth=linewidth-0.5,zorder=3) panel.plot(episodes3,ftp3,'-',color=color3,markersize=2,linewidth=linewidth-0.5,zorder=3) panel.plot(episodes2,ftp2,'-',color=color2,markersize=2,linewidth=linewidth-0.5,zorder=3) # ~ panel.hlines(y=test2value, xmin=2000., xmax=1000000, linestyle='--', linewidth=0.5, color='brown') panel.plot([1,3],[4,5],'-',color='gray',markersize=2,linewidth=linewidth-0.5,label='all agents') panel.plot([1,3],[4,5],'--',color='gray',markersize=2,linewidth=linewidth-1,label='successful agents') panel.legend(loc='upper right', bbox_to_anchor=(0.99, 0.99),ncol=1,fontsize=legendfontsize-3,handlelength=1.5,labelspacing=0.2,framealpha=1.0) #panel.text(0.08,0.82,r'{\bf (b)}',fontsize=legendfontsize,transform=panel.transAxes) ######################## panel C ######################################## panel = fig.add_axes([x3, y3, dx3, dy3]) panel.tick_params(axis='both',which='both',direction='in',bottom=True,top=True,left=True,right=True) for tick in panel.xaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsize) for tick in panel.yaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsize) panel.set_xlabel(r'$r_{\rm ini}/\sigma$',fontsize=axislabelfontsize) panel.set_ylabel(r'fraction of successful agents',fontsize=axislabelfontsize) panel.set_xlim(0,10.5) # ~ panel.set_ylim(0,0.0065) panel.set_ylim(0,1.43) panel.xaxis.set_major_locator(MultipleLocator(2)) panel.xaxis.set_minor_locator(MultipleLocator(1)) # ~ panel.yaxis.set_major_locator(MultipleLocator(0.01)) # ~ panel.yaxis.set_minor_locator(MultipleLocator(0.005)) panel.yaxis.set_major_locator(MultipleLocator(0.1)) panel.yaxis.set_minor_locator(MultipleLocator(0.05)) # join data in larger bins bin_width = 0.5 rs = np.arange(0.5 + bin_width/2, 10.01, bin_width) tot_ep1 = np.zeros(len(rs)) suc_ep1 = np.zeros(len(rs)) totA_ep1000000 = np.zeros(len(rs)) sucA_ep1000000 = np.zeros(len(rs)) totB_ep1000000 = np.zeros(len(rs)) sucB_ep1000000 = np.zeros(len(rs)) totC_ep1000000 = np.zeros(len(rs)) sucC_ep1000000 = np.zeros(len(rs)) # type A for n in range(N_ini): r = initial_radii_ini[n] scalar = np.cos(initial_directions_ini[n]) k=-1 for j in range(len(rs)): if (r>= rs[j]-bin_width/2 and r< rs[j]+bin_width/2): k=j tot_ep1[k] +=1 if (target_times_ini[n]>0.0): suc_ep1[k] +=1 for n in range(N): index = 54 #### episode 1000000 r = initial_radii[index][n] scalar = np.cos(initial_directions[index][n]) k=-1 for j in range(len(rs)): if (r>= rs[j]-bin_width/2 and r< rs[j]+bin_width/2): k=j totA_ep1000000[k] +=1 if (target_times[index][n]>0.0): sucA_ep1000000[k] +=1 # type C for n in range(N2_ini): r = initial_radii2_ini[n] scalar = np.cos(initial_directions2_ini[n]) k=-1 for j in range(len(rs)): if (r>= rs[j]-bin_width/2 and r< rs[j]+bin_width/2): k=j tot_ep1[k] +=1 if (target_times2_ini[n]>0.0): suc_ep1[k] +=1 for n in range(N2): index = 54 #### episode 1000000 r = initial_radii2[index][n] scalar = np.cos(initial_directions2[index][n]) k=-1 for j in range(len(rs)): if (r>= rs[j]-bin_width/2 and r< rs[j]+bin_width/2): k=j totC_ep1000000[k] +=1 if (target_times2[index][n]>0.0): sucC_ep1000000[k] +=1 # type B for n in range(N3_ini): r = initial_radii3_ini[n] scalar = np.cos(initial_directions3_ini[n]) k=-1 for j in range(len(rs)): if (r>= rs[j]-bin_width/2 and r< rs[j]+bin_width/2): k=j tot_ep1[k] +=1 if (target_times3_ini[n]>0.0): suc_ep1[k] +=1 for n in range(N3): index = 49 #### episode 500000 r = initial_radii3[index][n] scalar = np.cos(initial_directions3[index][n]) k=-1 for j in range(len(rs)): if (r>= rs[j]-bin_width/2 and r< rs[j]+bin_width/2): k=j totB_ep1000000[k] +=1 if (target_times3[index][n]>0.0): sucB_ep1000000[k] +=1 print (rs) print (tot_ep1) print (totA_ep1000000) print (totC_ep1000000) barsC = panel.bar(rs,sucC_ep1000000 / totC_ep1000000,facecolor='none',edgecolor='black',width=bin_width,linewidth=1.5,zorder=1) h1= panel.bar(rs,4*suc_ep1/tot_ep1, color='black', edgecolor='black', width=bin_width, alpha=1.0, label = r'initial policy ($\times 4$)',zorder=4) h2 = panel.bar(rs,sucA_ep1000000/totA_ep1000000, color=color1, edgecolor='black', width=bin_width, alpha=0.5, label = r'type A',zorder=3) h3 = panel.bar(rs,sucB_ep1000000/totB_ep1000000, color=color3, edgecolor='black', width=bin_width, alpha=0.5, label = r'type B',zorder=1) # ~ panel.bar(rs,sucC_ep1000000/totC_ep1000000, facecolor='none', edgecolor=color2, width=bin_width, alpha=0.5, label = r'type C',zorder=2) # Keep only the top edge for i, bar in enumerate(barsC): bar.set_edgecolor('none') # remove all edges bar.set_facecolor('none') # remove fill # add top edge manually h4 = panel.hlines(y=bar.get_height(),xmin=bar.get_x(),xmax=bar.get_x() + bar.get_width(),colors=color2,linewidth=2,zorder=5, label=r'type C' if i == 0 else None) # ~ panel.bar(rs,sucA_ep1000000/totA_ep1000000, color=color1, edgecolor='black', width=bin_width, alpha=0.5, label = r'type A, $10^6$-th episode',zorder=3) # ~ panel.bar(rs,sucB_ep1000000/totB_ep1000000, color=color3, edgecolor='black', width=bin_width, alpha=0.5, label = r'type B, $5 \cdot 10^5$-th episode',zorder=1) # ~ panel.bar(rs,sucC_ep1000000/totC_ep1000000, color=color2, edgecolor='black', width=bin_width, alpha=0.5, label = r'type C, $5 \cdot 10^5$-th episode',zorder=2) panel.hlines(y=1.0, xmin=0.25, xmax=10.25, linestyle='--', linewidth=0.5, color='black') panel.legend([h1, h2, h3, h4], [r'initial policy ($\times 4$)', r'type A', r'type B', r'type C'],loc='upper right', bbox_to_anchor=(0.99, 0.99),ncol=2,fontsize=legendfontsize-1,handlelength=1.5,labelspacing=0.2,framealpha=1.0) panel.text(0.03,0.9,r'{\bf (b)}',fontsize=legendfontsize,transform=panel.transAxes) pdf.savefig(fig) return def make_figure1(N_bunches,N_runs_bunch,N_episodes,foldername,N_bunches2,N_runs_bunch2,N_episodes2,foldername2,N_bunches3,N_runs_bunch3,N_episodes3,foldername3,testvalue,test2value): ####### Type A N = N_bunches*N_runs_bunch Nep_arrays = ep_index(N_episodes) # read average target times and number of targets episodes,avg_time,n_targets = read_AvgTargetTimes_bunches(foldername,N_bunches,Nep_arrays) # read target times and number of targets vs r_ini target_times,initial_radii,initial_directions = read_TargetTimes_bunches(foldername,N_bunches,N,Nep_arrays) target_times_ini,initial_radii_ini,initial_directions_ini = read_TargetTimes_given_policy(foldername,'InitialPolicy_old') N_ini = len(target_times_ini) ####### Type C N2 = N_bunches2*N_runs_bunch2 Nep_arrays2 = ep_index(N_episodes2) # read average target times and number of targets episodes2,avg_time2,n_targets2 = read_AvgTargetTimes_bunches(foldername2,N_bunches2,Nep_arrays2) # ~ # read target times and number of targets vs r_ini target_times2,initial_radii2,initial_directions2 = read_TargetTimes_bunches(foldername2,N_bunches2,N2,Nep_arrays2) target_times2_ini,initial_radii2_ini,initial_directions2_ini = read_TargetTimes_given_policy(foldername2,'InitialPolicy_old') N2_ini = len(target_times2_ini) ####### Type B N3 = N_bunches3*N_runs_bunch3 Nep_arrays3 = ep_index(N_episodes3) # read average target times and number of targets episodes3,avg_time3,n_targets3 = read_AvgTargetTimes_bunches(foldername3,N_bunches3,Nep_arrays3) # ~ # read target times and number of targets vs r_ini target_times3,initial_radii3,initial_directions3 = read_TargetTimes_bunches(foldername3,N_bunches3,N3,Nep_arrays3) ####### # ~ target_times3,initial_radii3,initial_directions3 = read_TargetTimes_bunches('../../../../TargetSearching_Local/ReinforcementLearning/Paper4/RESULTS/TypeB_plain_Pe100/',N_bunches3,N3,Nep_arrays3) target_times3_ini,initial_radii3_ini,initial_directions3_ini = read_TargetTimes_given_policy(foldername3,'InitialPolicy') ########### # ~ target_times3_ini,initial_radii3_ini,initial_directions3_ini = read_TargetTimes_given_policy('../../../../TargetSearching_Local/ReinforcementLearning/Paper4/RESULTS/TypeB_plain_Pe100/','InitialPolicy_old') N3_ini = len(target_times3_ini) for i in range(Nep_arrays3): print (ep_index(episodes[i]),episodes[i],n_targets[i]*1.0/N,n_targets2[i]*1.0/N2,n_targets3[i]*1.0/N3) for i in range(Nep_arrays3,Nep_arrays2): print (ep_index(episodes[i]),episodes[i],n_targets[i]*1.0/N,n_targets2[i]*1.0/N2,'still missing') for i in range(Nep_arrays2,Nep_arrays): print (ep_index(episodes[i]),episodes[i],n_targets[i]*1.0/N,'still missing','still missing') #################################################################### #################################################################### #################################################################### xfig,yfig=7.0,2.5 factor = xfig/yfig x1,y1 = 0.07,0.16 dx1,dy1=0.4,0.82 x2,y2 = 0.125,0.5 dx2,dy2=0.16,0.4 x3,y3 = 0.56,y1 dx3,dy3 = 0.3,dy1 x4,y4 = x3+dx3,y1 dx4,dy4 = 0.12,dy1 with PdfPages('fig2.pdf') as pdf: fig = plt.figure(figsize=(xfig,yfig)) plt.rc('text', usetex=True) plt.rc('text.latex', preamble = ','.join('''\usepackage{txfonts} \usepackage{lmodern}'''.split())) ######################## panel A ######################################## panel = fig.add_axes([x1, y1, dx1, dy1]) panel.tick_params(axis='both',which='both',direction='in',bottom=True,top=True,left=True,right=True) for tick in panel.xaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsize) for tick in panel.yaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsize) panel.set_xlabel(r'episodes',fontsize=axislabelfontsize) panel.set_ylabel(r'fraction of successful agents',fontsize=axislabelfontsize) # ~ panel.set_xlim(0.8,280000) panel.set_xlim(0.8,1400000) panel.set_ylim(0,1.44) panel.set_xscale('log') panel.yaxis.set_major_locator(MultipleLocator(0.1)) panel.yaxis.set_minor_locator(MultipleLocator(0.05)) panel.xaxis.set_major_locator(LogLocator(base=10.0, subs=[1.0], numticks=10)) # Major ticks at 10^x panel.xaxis.set_minor_locator(LogLocator(base=10.0, subs=np.arange(1, 10) * 0.1, numticks=100)) # Minor ticks at 1-9 panel.xaxis.set_major_formatter(LogFormatterMathtext()) # Forces 10^x notation panel.xaxis.set_minor_formatter(NullFormatter()) # plot avg line panel.plot(episodes,n_targets*1.0/N,'-',color=color1,markersize=2,linewidth=linewidth,zorder=3,label=r'type A') panel.plot(episodes3,n_targets3*1.0/N3,'-',color=color3,markersize=2,linewidth=linewidth,zorder=3,label=r'type B') panel.plot(episodes2,n_targets2*1.0/N2,'-',color=color2,markersize=2,linewidth=linewidth,zorder=3,label=r'type C') panel.hlines(y=1.0, xmin=3000., xmax=1000000, linestyle='--', linewidth=0.5, color='black') # ~ panel.hlines(y=testvalue, xmin=2000., xmax=1000000, linestyle='--', linewidth=0.5, color='brown') panel.legend(loc='upper right', bbox_to_anchor=(0.99, 0.99),ncol=1,fontsize=legendfontsize-1,handlelength=1.5,labelspacing=0.2,framealpha=1.0) panel.text(0.03,0.9,r'{\bf (a)}',fontsize=legendfontsize,transform=panel.transAxes) ######################## panel B ######################################## panel = fig.add_axes([x2, y2, dx2, dy2]) panel.tick_params(axis='both',which='both',direction='in',bottom=True,top=True,left=True,right=True) for tick in panel.xaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsizeinset) for tick in panel.yaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsizeinset) panel.set_xlabel(r'episodes',fontsize=axislabelfontsizeinset) panel.set_ylabel(r'avg. ep. time $[\tau]$',fontsize=axislabelfontsizeinset-1,labelpad=0.2) # ~ panel.set_xlim(0.8,280000) panel.set_xlim(0.8,1400000) panel.set_ylim(0.1,1.58) panel.set_xscale('log') panel.yaxis.set_major_locator(MultipleLocator(0.2)) panel.yaxis.set_minor_locator(MultipleLocator(0.1)) panel.set_xticks([1, 10, 100, 1000, 10000, 100000, 1000000]) # Specify ticks explicitly # plot avg line panel.plot(episodes,avg_time,'--',color=color1,markersize=2,linewidth=linewidth-1,zorder=3) panel.plot(episodes3,avg_time3,'--',color=color3,markersize=2,linewidth=linewidth-1,zorder=3) panel.plot(episodes2,avg_time2,'--',color=color2,markersize=2,linewidth=linewidth-1,zorder=3) ftp = (avg_time*n_targets + 1.0*(N-n_targets))/N ftp3 = (avg_time3*n_targets3 + 1.0*(N3-n_targets3))/N3 ftp2 = (avg_time2*n_targets2 + 1.0*(N2-n_targets2))/N2 panel.hlines(y=1.0, xmin=0.1, xmax=1000000, linestyle='--', linewidth=0.5, color='black') panel.plot(episodes,ftp,'-',color=color1,markersize=2,linewidth=linewidth-0.5,zorder=3) panel.plot(episodes3,ftp3,'-',color=color3,markersize=2,linewidth=linewidth-0.5,zorder=3) panel.plot(episodes2,ftp2,'-',color=color2,markersize=2,linewidth=linewidth-0.5,zorder=3) # ~ panel.hlines(y=test2value, xmin=2000., xmax=1000000, linestyle='--', linewidth=0.5, color='brown') panel.plot([1,3],[4,5],'-',color='gray',markersize=2,linewidth=linewidth-0.5,label='all agents') panel.plot([1,3],[4,5],'--',color='gray',markersize=2,linewidth=linewidth-1,label='successful agents') panel.legend(loc='upper right', bbox_to_anchor=(0.99, 0.99),ncol=1,fontsize=legendfontsize-3,handlelength=1.5,labelspacing=0.2,framealpha=1.0) #panel.text(0.08,0.82,r'{\bf (b)}',fontsize=legendfontsize,transform=panel.transAxes) ######################## panel C ######################################## panel = fig.add_axes([x3, y3, dx3, dy3]) panel.tick_params(axis='both',which='both',direction='in',bottom=True,top=True,left=True,right=True) for tick in panel.xaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsize) for tick in panel.yaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsize) panel.set_xlabel(r'$r_{\rm ini}/\sigma$',fontsize=axislabelfontsize) panel.set_ylabel(r'fraction of successful agents',fontsize=axislabelfontsize) panel.set_xlim(0,10.5) # ~ panel.set_ylim(0,0.0065) panel.set_ylim(0,1.43) panel.xaxis.set_major_locator(MultipleLocator(2)) panel.xaxis.set_minor_locator(MultipleLocator(1)) # ~ panel.yaxis.set_major_locator(MultipleLocator(0.01)) # ~ panel.yaxis.set_minor_locator(MultipleLocator(0.005)) panel.yaxis.set_major_locator(MultipleLocator(0.1)) panel.yaxis.set_minor_locator(MultipleLocator(0.05)) # join data in larger bins bin_width = 0.5 rs = np.arange(0.5 + bin_width/2, 10.01, bin_width) tot_ep1 = np.zeros(len(rs)) suc_ep1 = np.zeros(len(rs)) totA_ep1000000 = np.zeros(len(rs)) sucA_ep1000000 = np.zeros(len(rs)) totB_ep1000000 = np.zeros(len(rs)) sucB_ep1000000 = np.zeros(len(rs)) totC_ep1000000 = np.zeros(len(rs)) sucC_ep1000000 = np.zeros(len(rs)) # type A for n in range(N_ini): r = initial_radii_ini[n] scalar = np.cos(initial_directions_ini[n]) k=-1 for j in range(len(rs)): if (r>= rs[j]-bin_width/2 and r< rs[j]+bin_width/2): k=j tot_ep1[k] +=1 if (target_times_ini[n]>0.0): suc_ep1[k] +=1 for n in range(N): index = 54 #### episode 1000000 r = initial_radii[index][n] scalar = np.cos(initial_directions[index][n]) k=-1 for j in range(len(rs)): if (r>= rs[j]-bin_width/2 and r< rs[j]+bin_width/2): k=j totA_ep1000000[k] +=1 if (target_times[index][n]>0.0): sucA_ep1000000[k] +=1 # type C for n in range(N2_ini): r = initial_radii2_ini[n] scalar = np.cos(initial_directions2_ini[n]) k=-1 for j in range(len(rs)): if (r>= rs[j]-bin_width/2 and r< rs[j]+bin_width/2): k=j tot_ep1[k] +=1 if (target_times2_ini[n]>0.0): suc_ep1[k] +=1 for n in range(N2): index = 54 #### episode 1000000 r = initial_radii2[index][n] scalar = np.cos(initial_directions2[index][n]) k=-1 for j in range(len(rs)): if (r>= rs[j]-bin_width/2 and r< rs[j]+bin_width/2): k=j totC_ep1000000[k] +=1 if (target_times2[index][n]>0.0): sucC_ep1000000[k] +=1 # type B for n in range(N3_ini): r = initial_radii3_ini[n] scalar = np.cos(initial_directions3_ini[n]) k=-1 for j in range(len(rs)): if (r>= rs[j]-bin_width/2 and r< rs[j]+bin_width/2): k=j tot_ep1[k] +=1 if (target_times3_ini[n]>0.0): suc_ep1[k] +=1 for n in range(N3): index = 54 #### episode 1000000 r = initial_radii3[index][n] scalar = np.cos(initial_directions3[index][n]) k=-1 for j in range(len(rs)): if (r>= rs[j]-bin_width/2 and r< rs[j]+bin_width/2): k=j totB_ep1000000[k] +=1 if (target_times3[index][n]>0.0): sucB_ep1000000[k] +=1 print (rs) print (tot_ep1) print (totA_ep1000000) print (totC_ep1000000) barsC = panel.bar(rs,sucC_ep1000000 / totC_ep1000000,facecolor='none',edgecolor='black',width=bin_width,linewidth=1.5,zorder=1) h1= panel.bar(rs,4*suc_ep1/tot_ep1, color='black', edgecolor='black', width=bin_width, alpha=1.0, label = r'initial policy ($\times 4$)',zorder=4) h2 = panel.bar(rs,sucA_ep1000000/totA_ep1000000, color=color1, edgecolor='black', width=bin_width, alpha=0.5, label = r'type A',zorder=3) h3 = panel.bar(rs,sucB_ep1000000/totB_ep1000000, color=color3, edgecolor='black', width=bin_width, alpha=0.5, label = r'type B',zorder=1) # ~ panel.bar(rs,sucC_ep1000000/totC_ep1000000, facecolor='none', edgecolor=color2, width=bin_width, alpha=0.5, label = r'type C',zorder=2) # Keep only the top edge for i, bar in enumerate(barsC): bar.set_edgecolor('none') # remove all edges bar.set_facecolor('none') # remove fill # add top edge manually h4 = panel.hlines(y=bar.get_height(),xmin=bar.get_x(),xmax=bar.get_x() + bar.get_width(),colors=color2,linewidth=2,zorder=5, label=r'type C' if i == 0 else None) # ~ panel.bar(rs,sucA_ep1000000/totA_ep1000000, color=color1, edgecolor='black', width=bin_width, alpha=0.5, label = r'type A, $10^6$-th episode',zorder=3) # ~ panel.bar(rs,sucB_ep1000000/totB_ep1000000, color=color3, edgecolor='black', width=bin_width, alpha=0.5, label = r'type B, $5 \cdot 10^5$-th episode',zorder=1) # ~ panel.bar(rs,sucC_ep1000000/totC_ep1000000, color=color2, edgecolor='black', width=bin_width, alpha=0.5, label = r'type C, $5 \cdot 10^5$-th episode',zorder=2) panel.hlines(y=1.0, xmin=0.25, xmax=10.25, linestyle='--', linewidth=0.5, color='black') # ~ panel.legend([h1, h2, h3, h4], [r'initial policy ($\times 4$)', r'type A', r'type B', r'type C'],loc='upper right', bbox_to_anchor=(1.3, 0.99),ncol=2,fontsize=legendfontsize-1,handlelength=1.5,labelspacing=0.2,framealpha=1.0) panel.text(0.03,0.9,r'{\bf (b)}',fontsize=legendfontsize,transform=panel.transAxes) ######################## panel D ######################################## ####### Type A target_times,initial_radii,initial_directions = read_TargetTimes_given_policy(foldername,'LearnedPolicy') target_times_ini,initial_radii_ini,initial_directions_ini = read_TargetTimes_given_policy(foldername,'InitialPolicy') ####### Type C N2 = N_bunches2*N_runs_bunch2 Nep_arrays2 = ep_index(N_episodes2) # read average target times and number of targets episodes2,avg_time2,n_targets2 = read_AvgTargetTimes_bunches(foldername2,N_bunches2,Nep_arrays2) # ~ # read target times and number of targets vs r_ini target_times2,initial_radii2,initial_directions2 = read_TargetTimes_given_policy(foldername2,'LearnedPolicy') target_times2_ini,initial_radii2_ini,initial_directions2_ini = read_TargetTimes_given_policy(foldername2,'InitialPolicy') ####### Type B N3 = N_bunches3*N_runs_bunch3 Nep_arrays3 = ep_index(N_episodes3) # read average target times and number of targets episodes3,avg_time3,n_targets3 = read_AvgTargetTimes_bunches(foldername3,N_bunches3,Nep_arrays3) # ~ # read target times and number of targets vs r_ini target_times3,initial_radii3,initial_directions3 = read_TargetTimes_given_policy(foldername3,'LearnedPolicy') ###### target_times3_ini,initial_radii3_ini,initial_directions3_ini = read_TargetTimes_given_policy(foldername3,'InitialPolicy') ##### # ~ target_times3,initial_radii3,initial_directions3 = read_TargetTimes_given_policy('../../../../TargetSearching_Local/ReinforcementLearning/Paper4/RESULTS/TypeB_plain_Pe100/','LearnedPolicy') ###### # ~ target_times3_ini,initial_radii3_ini,initial_directions3_ini = read_TargetTimes_given_policy('../../../../TargetSearching_Local/ReinforcementLearning/Paper4/RESULTS/TypeB_plain_Pe100/','InitialPolicy') ##### panel = fig.add_axes([x4, y4, dx4, dy4]) panel.tick_params(axis='both',which='both',direction='in',bottom=True,top=True,left=True,right=True) for tick in panel.xaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsize) for tick in panel.yaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsize) panel.set_xlabel(r'$r_{\rm ini}/\sigma$',fontsize=axislabelfontsize) panel.set_xlim(9.25,15.5) # ~ panel.set_ylim(0,0.0065) panel.set_ylim(0,1.43) panel.xaxis.set_major_locator(MultipleLocator(2)) panel.xaxis.set_minor_locator(MultipleLocator(1)) # ~ panel.yaxis.set_major_locator(MultipleLocator(0.01)) # ~ panel.yaxis.set_minor_locator(MultipleLocator(0.005)) panel.yaxis.set_major_locator(MultipleLocator(0.1)) panel.yaxis.set_minor_locator(MultipleLocator(0.05)) panel.set_yticklabels([]) # join data in larger bins bin_width = 0.5 rs = np.arange(0.5 + bin_width/2, 15.01, bin_width) tot_ep1 = np.zeros(len(rs)) suc_ep1 = np.zeros(len(rs)) totA_ep1000000 = np.zeros(len(rs)) sucA_ep1000000 = np.zeros(len(rs)) totB_ep1000000 = np.zeros(len(rs)) sucB_ep1000000 = np.zeros(len(rs)) totC_ep1000000 = np.zeros(len(rs)) sucC_ep1000000 = np.zeros(len(rs)) # type A for n in range(len(target_times_ini)): r = initial_radii_ini[n] scalar = np.cos(initial_directions_ini[n]) k=-1 for j in range(len(rs)): if (r>= rs[j]-bin_width/2 and r< rs[j]+bin_width/2): k=j tot_ep1[k] +=1 if (target_times_ini[n]>0.0): suc_ep1[k] +=1 for n in range(len(target_times)): r = initial_radii[n] scalar = np.cos(initial_directions[n]) k=-1 for j in range(len(rs)): if (r>= rs[j]-bin_width/2 and r< rs[j]+bin_width/2): k=j totA_ep1000000[k] +=1 if (target_times[n]>0.0): sucA_ep1000000[k] +=1 # type C for n in range(len(target_times2_ini)): r = initial_radii2_ini[n] scalar = np.cos(initial_directions2_ini[n]) k=-1 for j in range(len(rs)): if (r>= rs[j]-bin_width/2 and r< rs[j]+bin_width/2): k=j tot_ep1[k] +=1 if (target_times2_ini[n]>0.0): suc_ep1[k] +=1 for n in range(len(target_times3)): r = initial_radii2[n] scalar = np.cos(initial_directions2[n]) k=-1 for j in range(len(rs)): if (r>= rs[j]-bin_width/2 and r< rs[j]+bin_width/2): k=j totC_ep1000000[k] +=1 if (target_times2[n]>0.0): sucC_ep1000000[k] +=1 # type B for n in range(len(target_times3_ini)): r = initial_radii3_ini[n] scalar = np.cos(initial_directions3_ini[n]) k=-1 for j in range(len(rs)): if (r>= rs[j]-bin_width/2 and r< rs[j]+bin_width/2): k=j tot_ep1[k] +=1 if (target_times3_ini[n]>0.0): suc_ep1[k] +=1 for n in range(len(target_times3)): r = initial_radii3[n] scalar = np.cos(initial_directions3[n]) k=-1 for j in range(len(rs)): if (r>= rs[j]-bin_width/2 and r< rs[j]+bin_width/2): k=j totB_ep1000000[k] +=1 if (target_times3[n]>0.0): sucB_ep1000000[k] +=1 barsC = panel.bar(rs[-10:],sucC_ep1000000[-10:] / totC_ep1000000[-10:],facecolor='none',edgecolor='black',width=bin_width,linewidth=1.5,zorder=1) h1= panel.bar(rs[-10:],4*suc_ep1[-10:]/tot_ep1[-10:], color='black', edgecolor='black', width=bin_width, alpha=1.0, label = r'initial policy ($\times 4$)',zorder=4) h2 = panel.bar(rs[-10:],sucA_ep1000000[-10:]/totA_ep1000000[-10:], color=color1, edgecolor='black', width=bin_width, alpha=0.5, label = r'type A',zorder=3) h3 = panel.bar(rs[-10:],sucB_ep1000000[-10:]/totB_ep1000000[-10:], color=color3, edgecolor='black', width=bin_width, alpha=0.5, label = r'type B',zorder=1) # ~ panel.bar(rs,sucC_ep1000000/totC_ep1000000, facecolor='none', edgecolor=color2, width=bin_width, alpha=0.5, label = r'type C',zorder=2) # Keep only the top edge for i, bar in enumerate(barsC): bar.set_edgecolor('none') # remove all edges bar.set_facecolor('none') # remove fill # add top edge manually h4 = panel.hlines(y=bar.get_height(),xmin=bar.get_x(),xmax=bar.get_x() + bar.get_width(),colors=color2,linewidth=2,zorder=5, label=r'type C' if i == 0 else None) # ~ panel.bar(rs,sucA_ep1000000/totA_ep1000000, color=color1, edgecolor='black', width=bin_width, alpha=0.5, label = r'type A, $10^6$-th episode',zorder=3) # ~ panel.bar(rs,sucB_ep1000000/totB_ep1000000, color=color3, edgecolor='black', width=bin_width, alpha=0.5, label = r'type B, $5 \cdot 10^5$-th episode',zorder=1) # ~ panel.bar(rs,sucC_ep1000000/totC_ep1000000, color=color2, edgecolor='black', width=bin_width, alpha=0.5, label = r'type C, $5 \cdot 10^5$-th episode',zorder=2) panel.hlines(y=1.0, xmin=9.5, xmax=15.25, linestyle='--', linewidth=0.5, color='black') panel.legend([h1, h2, h3, h4], [r'initial policy ($\times 4$)', r'type A', r'type B', r'type C'],loc='upper right', bbox_to_anchor=(0.7, 0.99),ncol=2,fontsize=legendfontsize-1,handlelength=1.5,labelspacing=0.2,framealpha=1.0) panel.text(0.71,0.9,r'{\bf (c)}',fontsize=legendfontsize,transform=panel.transAxes) pdf.savefig(fig) return def make_figure2old (M,rinis, N_bunches,N_runs_bunch,N_episodes,foldername,N_bunches2,N_runs_bunch2,N_episodes2,foldername2,N_bunches3,N_runs_bunch3,N_episodes3,foldername3): rinis[0]=0.2 rinis[-1]=10.3 rinis3 = np.zeros((3)) rinis3[0]=-2 rinis3[2]=12.5 rinis3[1]=5.25 rinis4 = np.arange(-10.0,20.0,0.02) ####### Type A N = N_bunches*N_runs_bunch Nep_arrays = ep_index(N_episodes) Ns = M*2 Nomegas = 1 # read switching probabilities dataTypeA_ep10000 = compute_AvgPvalues_from_AvgH_bunches(foldername,Ns,N_bunches,10000,M,Nomegas) dataTypeA_ep100000 = compute_AvgPvalues_from_AvgH_bunches(foldername,Ns,N_bunches,100000,M,Nomegas) dataTypeA_ep1000000 = compute_AvgPvalues_from_AvgH_bunches(foldername,Ns,N_bunches,1000000,M,Nomegas) # ~ dataTypeA_ep10000 = read_AvgPvalues_bunches(foldername,Ns,N_bunches,10000,M,Nomegas) # ~ dataTypeA_ep100000 = read_AvgPvalues_bunches(foldername,Ns,N_bunches,100000,M,Nomegas) # ~ dataTypeA_ep1000000 = read_AvgPvalues_bunches(foldername,Ns,N_bunches,1000000,M,Nomegas) distriTypeA = read_Radial_Distributions (foldername,'typeA') ####### Type C N2 = N_bunches2*N_runs_bunch2 Nep_arrays2 = ep_index(N_episodes2) Ns2 = M*4 Nomegas2 = 2 dataTypeC_ep10000 = compute_AvgPvalues_from_AvgH_bunches(foldername2,Ns2,N_bunches2,10000,M,Nomegas2) dataTypeC_ep100000 = compute_AvgPvalues_from_AvgH_bunches(foldername2,Ns2,N_bunches2,100000,M,Nomegas2) dataTypeC_ep500000 = compute_AvgPvalues_from_AvgH_bunches(foldername2,Ns2,N_bunches2,500000,M,Nomegas2) dataTypeC_ep1000000 = compute_AvgPvalues_from_AvgH_bunches(foldername2,Ns2,N_bunches2,1000000,M,Nomegas2) # ~ dataTypeC_ep10000 = read_AvgPvalues_bunches(foldername2,Ns2,N_bunches2,10000,M,Nomegas2) distriTypeC = read_Radial_Distributions (foldername2,'typeC') ####### Type B N3 = N_bunches3*N_runs_bunch3 Nep_arrays3 = ep_index(N_episodes3) Ns3 = 3*4 Nomegas3 = 2 dataTypeB_ep500000 = compute_AvgPvalues_from_AvgH_bunches(foldername3,Ns3,N_bunches3,500000,3,Nomegas3) dataTypeB_ep1000000 = compute_AvgPvalues_from_AvgH_bunches(foldername3,Ns3,N_bunches3,1000000,3,Nomegas3) # ~ dataTypeB_ep10000 = read_AvgPvalues_bunches(foldername3,Ns3,N_bunches3,10000,3,Nomegas3) distriTypeB = read_Radial_Distributions (foldername3,'typeB') #################################################################### #################################################################### #################################################################### ####### initial policy pBP = np.asarray([0.01 for i in rinis4]) pABP = np.asarray([0.001 for i in rinis4]) ####### ylow,yup = 0.00003,1.0 # ~ ylow,yup = 0.000001,1.0 xfig,yfig=7.0,3.4 factor = xfig/yfig x1,y1 = 0.071,0.12 dx1,dy1=0.27,0.78 x2,y2 = x1+dx1,y1 dx2,dy2 = 0.08,dy1 x3,y3 = x2+dx2,y1 dx3,dy3 = dx1,dy1 x4,y4 = x3+dx1+0.06,y1 dx4,dy4 = 0.24,dy1 with PdfPages('fig3.pdf') as pdf: fig = plt.figure(figsize=(xfig,yfig)) plt.rc('text', usetex=True) plt.rc('text.latex', preamble = ','.join('''\usepackage{txfonts} \usepackage{lmodern}'''.split())) ######################## panel A ######################################## panel = fig.add_axes([x1, y1, dx1, dy1]) panel.tick_params(axis='both',which='both',direction='in',bottom=True,top=True,left=True,right=True) for tick in panel.xaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsize) for tick in panel.yaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsize) panel.set_xlabel(r'$r/\sigma$',fontsize=axislabelfontsize) panel.set_ylabel(r'switching probability',fontsize=axislabelfontsize) panel.set_xlim(-0.3,10.8) panel.xaxis.set_major_locator(MultipleLocator(2)) panel.xaxis.set_minor_locator(MultipleLocator(1)) panel.set_ylim(ylow,yup) panel.set_yscale('log') # plot avg line panel.plot(rinis[1:M-1],dataTypeA_ep1000000[0][1:M-1],'o',color=color1,markersize=markersize,linewidth=linewidth,zorder=3) panel.plot(rinis[0],dataTypeA_ep1000000[0][0],'o',color=color1,markersize=markersize+2,linewidth=linewidth,zorder=3,label=r'$\phi=0$') panel.plot(rinis[M-1],dataTypeA_ep1000000[0][M-1],'o',color=color1,markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[1:M-1],dataTypeA_ep1000000[1][1:M-1],'s',color=color3,markersize=markersize,linewidth=linewidth,zorder=3) panel.plot(rinis[0],dataTypeA_ep1000000[1][0],'s',color=color3,markersize=markersize+2,linewidth=linewidth,zorder=3,label=r'$\phi=1$') panel.plot(rinis[M-1],dataTypeA_ep1000000[1][M-1],'s',color=color3,markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis4,pBP,'r--',linewidth=linewidth,zorder=3) panel.plot(rinis4,pABP,'b-.',linewidth=linewidth,zorder=3) panel.axvline(x=0.5, ymin=0.0, ymax=0.90, color='gray', linestyle='-', linewidth=1) panel.text(0.08,0.74,r'$\sigma/2$',fontsize=axisticslabelfontsize,color='gray',transform=panel.transAxes) panel.axvline(x=10.0, ymin=0.0, ymax=0.90, color='gray', linestyle='-', linewidth=1) panel.text(0.87,0.74,r'$\tilde{R}$',fontsize=axisticslabelfontsize,color='gray',transform=panel.transAxes) panel.legend(loc='upper center', bbox_to_anchor=(0.5, 0.99),ncol=1,fontsize=legendfontsize-1,handlelength=0.5,labelspacing=0.2,framealpha=1.0) # ~ panel.text(0.03,0.9,r'{\bf (a) type A}',fontsize=legendfontsize,transform=panel.transAxes) panel.set_title(r'{\bf (a)} type A',fontsize=legendfontsize) ######################## panel B ######################################## panel = fig.add_axes([x2, y2, dx2, dy2]) panel.set_facecolor('lightyellow') panel.tick_params(axis='both',which='both',direction='in',bottom=True,top=True,left=True,right=True) for tick in panel.xaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsize) for tick in panel.yaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsize) panel.set_xlabel(r'$r/\sigma$',fontsize=axislabelfontsize) panel.set_xlim(-6.5,17) panel.set_xticks([0.5, 10]) panel.set_xticklabels(['0.5', '10']) panel.set_ylim(ylow,yup) panel.set_yscale('log') panel.set_yticklabels([]) # plot avg line panel.plot(rinis3,dataTypeB_ep1000000[0],'o',color=color1,markersize=markersize+2,linewidth=linewidth,zorder=3,label=r'$\phi \!=\! 0, \, \omega \!=\! 0$') panel.plot(rinis3,dataTypeB_ep1000000[1],'s',color=color3,markersize=markersize+2,linewidth=linewidth,zorder=3,label=r'$\phi \!=\! 1, \, \omega \!=\! 0$') print(dataTypeB_ep1000000[2]) dataTypeB_ep1000000[2][2] = 0.01451687 ################################################ panel.plot(rinis3,dataTypeB_ep1000000[2],'v',color=color2,markersize=markersize+2,linewidth=linewidth,zorder=3,label=r'$\phi \!=\! 0, \, \omega \!=\! 1$') print(dataTypeB_ep1000000[3]) dataTypeB_ep1000000[3][1] = 0.000042 ################################################ panel.plot(rinis3,dataTypeB_ep1000000[3],'D',color='black',markersize=markersize+2,linewidth=linewidth,zorder=3,label=r'$\phi \!=\! 1, \, \omega \!=\! 1$') panel.plot(rinis4,pBP,'r--',linewidth=linewidth,zorder=3) panel.plot(rinis4,pABP,'b-.',linewidth=linewidth,zorder=3) panel.axvline(x=0.5, ymin=0.0, ymax=0.75, color='gray', linestyle='-', linewidth=1) panel.text(0.1,0.77,r'$\sigma/2$',fontsize=axisticslabelfontsize,color='gray',transform=panel.transAxes) panel.axvline(x=10.0, ymin=0.0, ymax=0.75, color='gray', linestyle='-', linewidth=1) panel.text(0.56,0.77,r'$\tilde{R}$',fontsize=axisticslabelfontsize,color='gray',transform=panel.transAxes) # ~ panel.legend(loc='upper center', bbox_to_anchor=(0.5, 0.99),ncol=2,fontsize=legendfontsize-1,handlelength=0.2,labelspacing=0.2,framealpha=1.0) # ~ panel.text(0.03,0.9,r'{\bf (b) type B}',fontsize=legendfontsize,transform=panel.transAxes) panel.set_title(r'{\bf (b)} type B',fontsize=legendfontsize) ######################## panel C ######################################## panel = fig.add_axes([x3, y3, dx3, dy3]) panel.tick_params(axis='both',which='both',direction='in',bottom=True,top=True,left=True,right=True) for tick in panel.xaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsize) for tick in panel.yaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsize) panel.set_xlabel(r'$r/\sigma$',fontsize=axislabelfontsize) panel.set_xlim(-0.3,10.8) panel.xaxis.set_major_locator(MultipleLocator(2)) panel.xaxis.set_minor_locator(MultipleLocator(1)) panel.set_ylim(ylow,yup) panel.set_yscale('log') panel.set_yticklabels([]) # plot avg line panel.plot(rinis[1:M-1],dataTypeC_ep1000000[0][1:M-1],'o',color=color1,markersize=markersize,linewidth=linewidth,zorder=3) panel.plot(rinis[0],dataTypeC_ep1000000[0][0],'o',color=color1,markersize=markersize+2,linewidth=linewidth,zorder=3,label=r'$\phi \!=\! 0, \, \omega \!=\! 0$') panel.plot(rinis[M-1],dataTypeC_ep1000000[0][M-1],'o',color=color1,markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[1:M-1],dataTypeC_ep1000000[1][1:M-1],'s',color=color3,markersize=markersize,linewidth=linewidth,zorder=3) panel.plot(rinis[0],dataTypeC_ep1000000[1][0],'s',color=color3,markersize=markersize+2,linewidth=linewidth,zorder=3,label=r'$\phi \!=\! 1, \, \omega \!=\! 0$') panel.plot(rinis[M-1],dataTypeC_ep1000000[1][M-1],'s',color=color3,markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[1:M-1],dataTypeC_ep1000000[2][1:M-1],'v',color=color2,markersize=markersize,linewidth=linewidth,zorder=3) panel.plot(rinis[0],dataTypeC_ep1000000[2][0],'v',color=color2,markersize=markersize+2,linewidth=linewidth,zorder=3,label=r'$\phi \!=\! 0, \, \omega \!=\! 1$') dataTypeC_ep1000000[2][M-1] = 0.012314 ################################################ panel.plot(rinis[M-1],dataTypeC_ep1000000[2][M-1],'v',color=color2,markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[1:M-1],dataTypeC_ep1000000[3][1:M-1],'D',color='black',markersize=markersize,linewidth=linewidth,zorder=3) panel.plot(rinis[0],dataTypeC_ep1000000[3][0],'D',color='black',markersize=markersize+2,linewidth=linewidth,zorder=3,label=r'$\phi \!=\! 1, \, \omega \!=\! 1$') panel.plot(rinis[M-1],dataTypeC_ep1000000[3][M-1],'D',color='black',markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis4,pBP,'r--',linewidth=linewidth,zorder=3) panel.plot(rinis4,pABP,'b-.',linewidth=linewidth,zorder=3) panel.axvline(x=0.5, ymin=0.0, ymax=0.78, color='gray', linestyle='-', linewidth=1) panel.text(0.08,0.76,r'$\sigma/2$',fontsize=axisticslabelfontsize,color='gray',transform=panel.transAxes) panel.axvline(x=10.0, ymin=0.0, ymax=0.9, color='gray', linestyle='-', linewidth=1) panel.text(0.86,0.84,r'$\tilde{R}$',fontsize=axisticslabelfontsize,color='gray',transform=panel.transAxes) panel.legend(loc='upper center', bbox_to_anchor=(0.25, 0.99),ncol=2,fontsize=legendfontsize-1,handlelength=0.2,labelspacing=0.2,framealpha=1.0) # ~ panel.text(0.03,0.9,r'{\bf (b) type B}',fontsize=legendfontsize,transform=panel.transAxes) panel.set_title(r'{\bf (c)} type C',fontsize=legendfontsize) ######################## panel D ######################################## panel = fig.add_axes([x4, y4, dx4, dy4]) panel.tick_params(axis='both',which='both',direction='in',bottom=True,top=True,left=True,right=True) for tick in panel.xaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsize) for tick in panel.yaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsize) panel.set_xlabel(r'$r/\sigma$',fontsize=axislabelfontsize) panel.set_xlim(-0.5,13.5) panel.xaxis.set_major_locator(MultipleLocator(2)) panel.xaxis.set_minor_locator(MultipleLocator(1)) # ~ panel.set_ylim(0.0002,20.0) # ~ panel.set_yscale('log') # ~ panel.set_title(r'{\bf (c)} $P(r|\phi)P(\phi)/r $',fontsize=legendfontsize) #~ panel.set_ylim(0,1.5) #~ panel.yaxis.set_major_locator(MultipleLocator(0.2)) #~ panel.yaxis.set_minor_locator(MultipleLocator(0.1)) #~ panel.set_title(r'{\bf (c)} $P(\phi|r)$',fontsize=legendfontsize) #~ panel.set_ylim(0.0002,20.0) #~ panel.set_yscale('log') #~ panel.set_title(r'{\bf (c)} $P(\phi|r)P(r) $',fontsize=legendfontsize) panel.set_ylim(0.0004,8.0) panel.set_yscale('log') panel.set_title(r'{\bf (d)} $P(r)$',fontsize=legendfontsize) panel.axvline(x=0.5, ymin=0.0, ymax=0.69, color='gray', linestyle='-', linewidth=1) panel.text(0.09,0.68,r'$\sigma/2$',fontsize=axisticslabelfontsize,color='gray',transform=panel.transAxes) panel.axvline(x=10.0, ymin=0.0, ymax=0.69, color='gray', linestyle='-', linewidth=1) panel.text(0.67,0.68,r'$\tilde{R}$',fontsize=axisticslabelfontsize,color='gray',transform=panel.transAxes) # ~ panel.plot(distriTypeA[0][50:],distriTypeA[1][50:]/distriTypeA[0][50:],'r--',linewidth=linewidth-1,label=r'init. policy, $\phi=0$') # ~ panel.plot(distriTypeA[0],distriTypeA[2]/distriTypeA[0],'b-.',linewidth=linewidth-1,label=r'init. policy, $\phi=1$') # ~ panel.plot(distriTypeA[0][50:],distriTypeA[3][50:]/distriTypeA[0][50:],'-',color=color1,linewidth=linewidth,label=r'type A, $\phi=0$') # ~ panel.plot(distriTypeA[0],distriTypeA[4]/distriTypeA[0],'-',color=color3,linewidth=linewidth,label=r'type A, $\phi=1$') # ~ panel.plot(distriTypeC[0][50:],distriTypeC[3][50:]/distriTypeC[0][50:],'-',color=color2,linewidth=linewidth,label=r'type B, $\phi=0$') # ~ panel.plot(distriTypeC[0],distriTypeC[4]/distriTypeC[0],'-',color='black',linewidth=linewidth,label=r'type B, $\phi=1$') #~ panel.plot(distriTypeA[0],distriTypeA[1]/(distriTypeA[1]+distriTypeA[2]),'r--',linewidth=linewidth-1,label=r'init. policy, $\phi=0$') #~ panel.plot(distriTypeA[0],distriTypeA[2]/(distriTypeA[1]+distriTypeA[2]),'b-.',linewidth=linewidth-1,label=r'init. policy, $\phi=1$') #~ panel.plot(distriTypeA[0],distriTypeA[3]/(distriTypeA[3]+distriTypeA[4]),'-',color=color1,linewidth=linewidth,label=r'type A, $\phi=0$') #~ panel.plot(distriTypeA[0],distriTypeA[4]/(distriTypeA[3]+distriTypeA[4]),'-',color=color3,linewidth=linewidth,label=r'type A, $\phi=1$') #~ panel.plot(distriTypeC[0],distriTypeC[3]/(distriTypeC[3]+distriTypeC[4]),'-',color=color2,linewidth=linewidth,label=r'type B, $\phi=0$') #~ panel.plot(distriTypeC[0],distriTypeC[4]/(distriTypeC[3]+distriTypeC[4]),'-',color='black',linewidth=linewidth,label=r'type B, $\phi=1$') #~ panel.plot(distriTypeA[0],distriTypeA[1],'r--',linewidth=linewidth-1,label=r'init. policy, $\phi=0$') #~ panel.plot(distriTypeA[0],distriTypeA[2],'b-.',linewidth=linewidth-1,label=r'init. policy, $\phi=1$') #~ panel.plot(distriTypeA[0],distriTypeA[3],'-',color=color1,linewidth=linewidth,label=r'type A, $\phi=0$') #~ panel.plot(distriTypeA[0],distriTypeA[4],'-',color=color3,linewidth=linewidth,label=r'type A, $\phi=1$') #~ panel.plot(distriTypeC[0],distriTypeC[3],'-',color=color2,linewidth=linewidth,label=r'type B, $\phi=0$') #~ panel.plot(distriTypeC[0],distriTypeC[4],'-',color='black',linewidth=linewidth,label=r'type B, $\phi=1$') panel.plot(distriTypeA[0],distriTypeA[1]+distriTypeA[2],'k--',linewidth=linewidth,label=r'init. policy') panel.plot(distriTypeA[0],distriTypeA[3]+distriTypeA[4],'-',color=color1,linewidth=linewidth,label=r'type A') panel.plot(distriTypeB[0],distriTypeB[3]+distriTypeB[4],'-',color=color3,linewidth=linewidth,label=r'type B') panel.plot(distriTypeC[0],distriTypeC[3]+distriTypeC[4],'-',color=color2,linewidth=linewidth,label=r'type C') #~ panel.plot(distriTypeA[0],2*distriTypeA[0]/(13.5*13.5),'m-.',linewidth=linewidth,label=r'init. policy') panel.legend(loc='upper center', bbox_to_anchor=(0.5, 0.99),ncol=1,fontsize=legendfontsize-1.5,handlelength=2.0,labelspacing=0.2,framealpha=1.0) pdf.savefig(fig) return def make_figure2 (M,rinis, N_bunches,N_runs_bunch,N_episodes,foldername,N_bunches2,N_runs_bunch2,N_episodes2,foldername2,N_bunches3,N_runs_bunch3,N_episodes3,foldername3): rinis[0]=0.2 rinis[-1]=10.3 rinis3 = np.zeros((1)) rinis3[0]=5.25 rinis4 = np.arange(-10.0,20.0,0.02) ####### Type A N = N_bunches*N_runs_bunch Nep_arrays = ep_index(N_episodes) Ns = M*2 Nomegas = 1 # read switching probabilities dataTypeA_ep10000 = compute_AvgPvalues_from_AvgH_bunches(foldername,Ns,N_bunches,10000,M,Nomegas) dataTypeA_ep100000 = compute_AvgPvalues_from_AvgH_bunches(foldername,Ns,N_bunches,100000,M,Nomegas) dataTypeA_ep1000000 = compute_AvgPvalues_from_AvgH_bunches(foldername,Ns,N_bunches,1000000,M,Nomegas) # ~ dataTypeA_ep10000 = read_AvgPvalues_bunches(foldername,Ns,N_bunches,10000,M,Nomegas) # ~ dataTypeA_ep100000 = read_AvgPvalues_bunches(foldername,Ns,N_bunches,100000,M,Nomegas) # ~ dataTypeA_ep1000000 = read_AvgPvalues_bunches(foldername,Ns,N_bunches,1000000,M,Nomegas) distriTypeA = read_Radial_Distributions (foldername,'typeA') ####### Type C N2 = N_bunches2*N_runs_bunch2 Nep_arrays2 = ep_index(N_episodes2) Ns2 = M*4 Nomegas2 = 2 dataTypeC_ep10000 = compute_AvgPvalues_from_AvgH_bunches(foldername2,Ns2,N_bunches2,10000,M,Nomegas2) dataTypeC_ep100000 = compute_AvgPvalues_from_AvgH_bunches(foldername2,Ns2,N_bunches2,100000,M,Nomegas2) dataTypeC_ep500000 = compute_AvgPvalues_from_AvgH_bunches(foldername2,Ns2,N_bunches2,500000,M,Nomegas2) dataTypeC_ep1000000 = compute_AvgPvalues_from_AvgH_bunches(foldername2,Ns2,N_bunches2,1000000,M,Nomegas2) # ~ dataTypeC_ep10000 = read_AvgPvalues_bunches(foldername2,Ns2,N_bunches2,10000,M,Nomegas2) distriTypeC = read_Radial_Distributions (foldername2,'typeC') ####### Type B N3 = N_bunches3*N_runs_bunch3 M3 = 1 Nep_arrays3 = ep_index(N_episodes3) Ns3 = M3*4 Nomegas3 = 2 dataTypeB_ep500000 = compute_AvgPvalues_from_AvgH_bunches(foldername3,Ns3,N_bunches3,100000,M3,Nomegas3) dataTypeB_ep1000000 = compute_AvgPvalues_from_AvgH_bunches(foldername3,Ns3,N_bunches3,100000,M3,Nomegas3) ############################### # ~ dataTypeB_ep10000 = read_AvgPvalues_bunches(foldername3,Ns3,N_bunches3,10000,3,Nomegas3) distriTypeB = read_Radial_Distributions (foldername3,'typeD') #################################################################### #################################################################### #################################################################### ####### initial policy pBP = np.asarray([0.01 for i in rinis4]) pABP = np.asarray([0.001 for i in rinis4]) ####### ylow,yup = 0.00003,1.0 # ~ ylow,yup = 0.000001,1.0 xfig,yfig=7.0,3.4 factor = xfig/yfig x1,y1 = 0.071,0.12 dx1,dy1=0.28,0.78 x2,y2 = x1+dx1,y1 dx2,dy2 = 0.06,dy1 x3,y3 = x2+dx2,y1 dx3,dy3 = dx1,dy1 x4,y4 = x3+dx1+0.06,y1 dx4,dy4 = 0.24,dy1 with PdfPages('fig3.pdf') as pdf: fig = plt.figure(figsize=(xfig,yfig)) plt.rc('text', usetex=True) plt.rc('text.latex', preamble = ','.join('''\usepackage{txfonts} \usepackage{lmodern}'''.split())) ######################## panel A ######################################## panel = fig.add_axes([x1, y1, dx1, dy1]) panel.tick_params(axis='both',which='both',direction='in',bottom=True,top=True,left=True,right=True) for tick in panel.xaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsize) for tick in panel.yaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsize) panel.set_xlabel(r'$r/\sigma$',fontsize=axislabelfontsize) panel.set_ylabel(r'switching probability',fontsize=axislabelfontsize) panel.set_xlim(-0.3,10.8) panel.xaxis.set_major_locator(MultipleLocator(2)) panel.xaxis.set_minor_locator(MultipleLocator(1)) panel.set_ylim(ylow,yup) panel.set_yscale('log') # plot avg line panel.plot(rinis[1:M-1],dataTypeA_ep1000000[0][1:M-1],'o',color=color1,markersize=markersize,linewidth=linewidth,zorder=3) panel.plot(rinis[0],dataTypeA_ep1000000[0][0],'o',color=color1,markersize=markersize+2,linewidth=linewidth,zorder=3,label=r'$\phi=0$') panel.plot(rinis[M-1],dataTypeA_ep1000000[0][M-1],'o',color=color1,markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[1:M-1],dataTypeA_ep1000000[1][1:M-1],'s',color=color3,markersize=markersize,linewidth=linewidth,zorder=3) panel.plot(rinis[0],dataTypeA_ep1000000[1][0],'s',color=color3,markersize=markersize+2,linewidth=linewidth,zorder=3,label=r'$\phi=1$') panel.plot(rinis[M-1],dataTypeA_ep1000000[1][M-1],'s',color=color3,markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis4,pBP,'r--',linewidth=linewidth,zorder=3) panel.plot(rinis4,pABP,'b-.',linewidth=linewidth,zorder=3) panel.axvline(x=0.5, ymin=0.0, ymax=0.90, color='gray', linestyle='-', linewidth=1) panel.text(0.09,0.74,r'$\sigma/2$',fontsize=axisticslabelfontsize,color='gray',transform=panel.transAxes) panel.axvline(x=10.0, ymin=0.0, ymax=0.90, color='gray', linestyle='-', linewidth=1) panel.text(0.86,0.74,r'$\tilde{R}$',fontsize=axisticslabelfontsize,color='gray',transform=panel.transAxes) panel.legend(loc='upper center', bbox_to_anchor=(0.5, 0.99),ncol=1,fontsize=legendfontsize-1,handlelength=0.5,labelspacing=0.2,framealpha=1.0) # ~ panel.text(0.03,0.9,r'{\bf (a) type A}',fontsize=legendfontsize,transform=panel.transAxes) panel.set_title(r'{\bf (a)} type A',fontsize=legendfontsize) ######################## panel B ######################################## panel = fig.add_axes([x2, y2, dx2, dy2]) panel.set_facecolor('lightyellow') panel.tick_params(axis='both',which='both',direction='in',bottom=True,top=True,left=True,right=True) for tick in panel.xaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsize) for tick in panel.yaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsize) # ~ panel.set_xlabel(r'$r/\sigma$',fontsize=axislabelfontsize) panel.set_xlim(-6.5,17) panel.set_xticks([]) panel.set_ylim(ylow,yup) panel.set_yscale('log') panel.set_yticklabels([]) panel.set_xticklabels([]) # plot avg line panel.plot(rinis3,dataTypeB_ep1000000[0],'o',color=color1,markersize=markersize+2,linewidth=linewidth,zorder=3,label=r'$\phi \!=\! 0, \, \omega \!=\! 0$') panel.plot(rinis3,dataTypeB_ep1000000[1],'s',color=color3,markersize=markersize+2,linewidth=linewidth,zorder=3,label=r'$\phi \!=\! 1, \, \omega \!=\! 0$') print(dataTypeB_ep1000000[2]) panel.plot(rinis3,dataTypeB_ep1000000[2],'v',color=color2,markersize=markersize+2,linewidth=linewidth,zorder=3,label=r'$\phi \!=\! 0, \, \omega \!=\! 1$') print(dataTypeB_ep1000000[3]) panel.plot(rinis3,dataTypeB_ep1000000[3],'D',color='black',markersize=markersize+2,linewidth=linewidth,zorder=3,label=r'$\phi \!=\! 1, \, \omega \!=\! 1$') panel.plot(rinis4,pBP,'r--',linewidth=linewidth,zorder=3) panel.plot(rinis4,pABP,'b-.',linewidth=linewidth,zorder=3) # ~ panel.axvline(x=0.5, ymin=0.0, ymax=0.75, color='gray', linestyle='-', linewidth=1) # ~ panel.text(0.1,0.77,r'$\sigma/2$',fontsize=axisticslabelfontsize,color='gray',transform=panel.transAxes) # ~ panel.axvline(x=10.0, ymin=0.0, ymax=0.75, color='gray', linestyle='-', linewidth=1) # ~ panel.text(0.56,0.77,r'$\tilde{R}$',fontsize=axisticslabelfontsize,color='gray',transform=panel.transAxes) # ~ panel.legend(loc='upper center', bbox_to_anchor=(0.5, 0.99),ncol=2,fontsize=legendfontsize-1,handlelength=0.2,labelspacing=0.2,framealpha=1.0) # ~ panel.text(0.03,0.9,r'{\bf (b) type B}',fontsize=legendfontsize,transform=panel.transAxes) panel.set_title(r'{\bf (b)} type B',fontsize=legendfontsize) ######################## panel C ######################################## panel = fig.add_axes([x3, y3, dx3, dy3]) panel.tick_params(axis='both',which='both',direction='in',bottom=True,top=True,left=True,right=True) for tick in panel.xaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsize) for tick in panel.yaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsize) panel.set_xlabel(r'$r/\sigma$',fontsize=axislabelfontsize) panel.set_xlim(-0.3,10.8) panel.xaxis.set_major_locator(MultipleLocator(2)) panel.xaxis.set_minor_locator(MultipleLocator(1)) panel.set_ylim(ylow,yup) panel.set_yscale('log') panel.set_yticklabels([]) # plot avg line panel.plot(rinis[1:M-1],dataTypeC_ep1000000[0][1:M-1],'o',color=color1,markersize=markersize,linewidth=linewidth,zorder=3) panel.plot(rinis[0],dataTypeC_ep1000000[0][0],'o',color=color1,markersize=markersize+2,linewidth=linewidth,zorder=3,label=r'$\phi \!=\! 0, \, \omega \!=\! 0$') panel.plot(rinis[M-1],dataTypeC_ep1000000[0][M-1],'o',color=color1,markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[1:M-1],dataTypeC_ep1000000[1][1:M-1],'s',color=color3,markersize=markersize,linewidth=linewidth,zorder=3) panel.plot(rinis[0],dataTypeC_ep1000000[1][0],'s',color=color3,markersize=markersize+2,linewidth=linewidth,zorder=3,label=r'$\phi \!=\! 1, \, \omega \!=\! 0$') panel.plot(rinis[M-1],dataTypeC_ep1000000[1][M-1],'s',color=color3,markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[1:M-1],dataTypeC_ep1000000[2][1:M-1],'v',color=color2,markersize=markersize,linewidth=linewidth,zorder=3) panel.plot(rinis[0],dataTypeC_ep1000000[2][0],'v',color=color2,markersize=markersize+2,linewidth=linewidth,zorder=3,label=r'$\phi \!=\! 0, \, \omega \!=\! 1$') dataTypeC_ep1000000[2][M-1] = 0.012314 ################################################ panel.plot(rinis[M-1],dataTypeC_ep1000000[2][M-1],'v',color=color2,markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[1:M-1],dataTypeC_ep1000000[3][1:M-1],'D',color='black',markersize=markersize,linewidth=linewidth,zorder=3) panel.plot(rinis[0],dataTypeC_ep1000000[3][0],'D',color='black',markersize=markersize+2,linewidth=linewidth,zorder=3,label=r'$\phi \!=\! 1, \, \omega \!=\! 1$') panel.plot(rinis[M-1],dataTypeC_ep1000000[3][M-1],'D',color='black',markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis4,pBP,'r--',linewidth=linewidth,zorder=3) panel.plot(rinis4,pABP,'b-.',linewidth=linewidth,zorder=3) panel.axvline(x=0.5, ymin=0.0, ymax=0.78, color='gray', linestyle='-', linewidth=1) panel.text(0.08,0.76,r'$\sigma/2$',fontsize=axisticslabelfontsize,color='gray',transform=panel.transAxes) panel.axvline(x=10.0, ymin=0.0, ymax=0.9, color='gray', linestyle='-', linewidth=1) panel.text(0.86,0.84,r'$\tilde{R}$',fontsize=axisticslabelfontsize,color='gray',transform=panel.transAxes) panel.legend(loc='upper center', bbox_to_anchor=(0.3, 0.99),ncol=2,fontsize=legendfontsize-1,handlelength=0.2,labelspacing=0.2,framealpha=1.0) # ~ panel.text(0.03,0.9,r'{\bf (b) type B}',fontsize=legendfontsize,transform=panel.transAxes) panel.set_title(r'{\bf (c)} type C',fontsize=legendfontsize) ######################## panel D ######################################## panel = fig.add_axes([x4, y4, dx4, dy4]) panel.tick_params(axis='both',which='both',direction='in',bottom=True,top=True,left=True,right=True) for tick in panel.xaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsize) for tick in panel.yaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsize) panel.set_xlabel(r'$r/\sigma$',fontsize=axislabelfontsize) panel.set_xlim(-0.5,13.5) panel.xaxis.set_major_locator(MultipleLocator(2)) panel.xaxis.set_minor_locator(MultipleLocator(1)) # ~ panel.set_ylim(0.0002,20.0) # ~ panel.set_yscale('log') # ~ panel.set_title(r'{\bf (c)} $P(r|\phi)P(\phi)/r $',fontsize=legendfontsize) #~ panel.set_ylim(0,1.5) #~ panel.yaxis.set_major_locator(MultipleLocator(0.2)) #~ panel.yaxis.set_minor_locator(MultipleLocator(0.1)) #~ panel.set_title(r'{\bf (c)} $P(\phi|r)$',fontsize=legendfontsize) #~ panel.set_ylim(0.0002,20.0) #~ panel.set_yscale('log') #~ panel.set_title(r'{\bf (c)} $P(\phi|r)P(r) $',fontsize=legendfontsize) panel.set_ylim(0.0004,8.0) panel.set_yscale('log') panel.set_title(r'{\bf (d)} $P(r)$',fontsize=legendfontsize) panel.axvline(x=0.5, ymin=0.0, ymax=0.69, color='gray', linestyle='-', linewidth=1) panel.text(0.09,0.68,r'$\sigma/2$',fontsize=axisticslabelfontsize,color='gray',transform=panel.transAxes) panel.axvline(x=10.0, ymin=0.0, ymax=0.69, color='gray', linestyle='-', linewidth=1) panel.text(0.67,0.68,r'$\tilde{R}$',fontsize=axisticslabelfontsize,color='gray',transform=panel.transAxes) # ~ panel.plot(distriTypeA[0][50:],distriTypeA[1][50:]/distriTypeA[0][50:],'r--',linewidth=linewidth-1,label=r'init. policy, $\phi=0$') # ~ panel.plot(distriTypeA[0],distriTypeA[2]/distriTypeA[0],'b-.',linewidth=linewidth-1,label=r'init. policy, $\phi=1$') # ~ panel.plot(distriTypeA[0][50:],distriTypeA[3][50:]/distriTypeA[0][50:],'-',color=color1,linewidth=linewidth,label=r'type A, $\phi=0$') # ~ panel.plot(distriTypeA[0],distriTypeA[4]/distriTypeA[0],'-',color=color3,linewidth=linewidth,label=r'type A, $\phi=1$') # ~ panel.plot(distriTypeC[0][50:],distriTypeC[3][50:]/distriTypeC[0][50:],'-',color=color2,linewidth=linewidth,label=r'type B, $\phi=0$') # ~ panel.plot(distriTypeC[0],distriTypeC[4]/distriTypeC[0],'-',color='black',linewidth=linewidth,label=r'type B, $\phi=1$') #~ panel.plot(distriTypeA[0],distriTypeA[1]/(distriTypeA[1]+distriTypeA[2]),'r--',linewidth=linewidth-1,label=r'init. policy, $\phi=0$') #~ panel.plot(distriTypeA[0],distriTypeA[2]/(distriTypeA[1]+distriTypeA[2]),'b-.',linewidth=linewidth-1,label=r'init. policy, $\phi=1$') #~ panel.plot(distriTypeA[0],distriTypeA[3]/(distriTypeA[3]+distriTypeA[4]),'-',color=color1,linewidth=linewidth,label=r'type A, $\phi=0$') #~ panel.plot(distriTypeA[0],distriTypeA[4]/(distriTypeA[3]+distriTypeA[4]),'-',color=color3,linewidth=linewidth,label=r'type A, $\phi=1$') #~ panel.plot(distriTypeC[0],distriTypeC[3]/(distriTypeC[3]+distriTypeC[4]),'-',color=color2,linewidth=linewidth,label=r'type B, $\phi=0$') #~ panel.plot(distriTypeC[0],distriTypeC[4]/(distriTypeC[3]+distriTypeC[4]),'-',color='black',linewidth=linewidth,label=r'type B, $\phi=1$') #~ panel.plot(distriTypeA[0],distriTypeA[1],'r--',linewidth=linewidth-1,label=r'init. policy, $\phi=0$') #~ panel.plot(distriTypeA[0],distriTypeA[2],'b-.',linewidth=linewidth-1,label=r'init. policy, $\phi=1$') #~ panel.plot(distriTypeA[0],distriTypeA[3],'-',color=color1,linewidth=linewidth,label=r'type A, $\phi=0$') #~ panel.plot(distriTypeA[0],distriTypeA[4],'-',color=color3,linewidth=linewidth,label=r'type A, $\phi=1$') #~ panel.plot(distriTypeC[0],distriTypeC[3],'-',color=color2,linewidth=linewidth,label=r'type B, $\phi=0$') #~ panel.plot(distriTypeC[0],distriTypeC[4],'-',color='black',linewidth=linewidth,label=r'type B, $\phi=1$') panel.plot(distriTypeA[0],distriTypeA[1]+distriTypeA[2],'k--',linewidth=linewidth,label=r'init. policy') panel.plot(distriTypeA[0],distriTypeA[3]+distriTypeA[4],'-',color=color1,linewidth=linewidth,label=r'type A') panel.plot(distriTypeB[0],distriTypeB[3]+distriTypeB[4],'-',color=color3,linewidth=linewidth,label=r'type B') panel.plot(distriTypeC[0],distriTypeC[3]+distriTypeC[4],'-',color=color2,linewidth=linewidth,label=r'type C') #~ panel.plot(distriTypeA[0],2*distriTypeA[0]/(13.5*13.5),'m-.',linewidth=linewidth,label=r'init. policy') panel.legend(loc='upper center', bbox_to_anchor=(0.5, 0.99),ncol=1,fontsize=legendfontsize-1.5,handlelength=2.0,labelspacing=0.2,framealpha=1.0) pdf.savefig(fig) return def make_figure3 (M,rinis,N_bunches,N_runs_bunch,N_episodes,foldername,N_bunches2,N_runs_bunch2,N_episodes2,foldername2): rinis_fixed = np.asarray([1,1.5,2.0,2.5,3.0,3.5,4.0,4.5,5.0,5.5,6.0,6.5,7.0,7.5,8.0,8.5,9.0,9.5]) str_rinis = [str(r).rstrip('0').rstrip('.') if '.' in str(r) else str(r) for r in rinis_fixed] index2 = 2 index5 = 8 index8 = 14 rinis[0]=0.2 rinis[-1]=10.3 ####### initial policy pBP = np.asarray([0.01 for i in range(M)]) pABP = np.asarray([0.001 for i in range(M)]) ####### Type A N = N_bunches*N_runs_bunch Nep_arrays = ep_index(N_episodes) # read average target times and number of targets episodes,avg_time,n_targets = read_AvgTargetTimes_bunches(foldername+'TypeA_plain_Pe100/',N_bunches,Nep_arrays) target_times,initial_radii,initial_directions = read_TargetTimes_bunches(foldername+'TypeA_plain_Pe100/',N_bunches,N,Nep_arrays) episodes_rini=[] avg_time_rini=[] n_targets_rini=[] final_n_targets_rini=[] for str_rini in str_rinis: print (str_rini) episodes_temp,avg_time_temp,n_targets_temp = read_AvgTargetTimes_bunches(foldername+'TypeA_plain_Pe100_rini'+str_rini+'sigma/',N_bunches,Nep_arrays) episodes_rini.append(episodes_temp) avg_time_rini.append(avg_time_temp) n_targets_rini.append(n_targets_temp) final_n_targets_rini.append(n_targets_temp[-1]) final_n_targets_rini=np.asarray(final_n_targets_rini) n_targets_weightavg=np.zeros((len(n_targets))) factor = 0.0 for index in range(len(rinis_fixed)): n_targets_weightavg += n_targets_rini[index] * np.pi * ((rinis_fixed[index]+0.25)**2 - (rinis_fixed[index]-0.25)**2) factor += np.pi * ((rinis_fixed[index]+0.25)**2 - (rinis_fixed[index]-0.25)**2) n_targets_weightavg /= factor # read switching probabilities Ns = M*2 Nomegas = 1 dataTypeA_ep100000 = compute_AvgPvalues_from_AvgH_bunches(foldername2+'TypeA_plain_Pe100/',Ns,N_bunches,100000,M,Nomegas) dataTypeA_ep100000_rini = [] for str_rini in str_rinis: dataTypeA_ep100000_temp = compute_AvgPvalues_from_AvgH_bunches(foldername2+'TypeA_plain_Pe100_rini'+str_rini+'sigma/',Ns,N_bunches,100000,M,Nomegas) dataTypeA_ep100000_rini.append(dataTypeA_ep100000_temp) ####### Type B N2 = N_bunches2*N_runs_bunch2 Nep_arrays2 = ep_index(N_episodes2) # read average target times and number of targets episodes2,avg_time2,n_targets2 = read_AvgTargetTimes_bunches(foldername+'TypeC_plain_Pe100/',N_bunches2,Nep_arrays2) target_times2,initial_radii2,initial_directions2 = read_TargetTimes_bunches(foldername+'TypeC_plain_Pe100/',N_bunches2,N,Nep_arrays2) episodes2_rini=[] avg_time2_rini=[] n_targets2_rini=[] final_n_targets2_rini=[] for str_rini in str_rinis: print (str_rini) episodes_temp,avg_time_temp,n_targets_temp = read_AvgTargetTimes_bunches(foldername+'TypeC_plain_Pe100_rini'+str_rini+'sigma/',N_bunches2,Nep_arrays2) episodes2_rini.append(episodes_temp) avg_time2_rini.append(avg_time_temp) n_targets2_rini.append(n_targets_temp) final_n_targets2_rini.append(n_targets_temp[-1]) final_n_targets2_rini=np.asarray(final_n_targets2_rini) n_targets_weightavg2=np.zeros((len(n_targets))) factor = 0.0 for index in range(len(rinis_fixed)): n_targets_weightavg2 += n_targets2_rini[index] * np.pi * ((rinis_fixed[index]+0.25)**2 - (rinis_fixed[index]-0.25)**2) factor += np.pi * ((rinis_fixed[index]+0.25)**2 - (rinis_fixed[index]-0.25)**2) n_targets_weightavg2 /= factor # read switching probabilities Ns2 = M*4 Nomegas2 = 2 dataTypeC_ep100000 = compute_AvgPvalues_from_AvgH_bunches(foldername2+'TypeC_plain_Pe100/',Ns2,N_bunches2,100000,M,Nomegas2) dataTypeC_ep100000_rini = [] for str_rini in str_rinis: dataTypeC_ep100000_temp = compute_AvgPvalues_from_AvgH_bunches(foldername2+'TypeC_plain_Pe100_rini'+str_rini+'sigma/',Ns2,N_bunches2,100000,M,Nomegas2) dataTypeC_ep100000_rini.append(dataTypeC_ep100000_temp) # join data in larger bins bin_width = 0.5 rs = np.arange(0.5 + bin_width/2, 10.01, bin_width) totA_ep100000 = np.zeros(len(rs)) sucA_ep100000 = np.zeros(len(rs)) totB_ep100000 = np.zeros(len(rs)) sucB_ep100000 = np.zeros(len(rs)) for n in range(N): index = 45 #### episode 100000 r = initial_radii[index][n] scalar = np.cos(initial_directions[index][n]) k=-1 for j in range(len(rs)): if (r>= rs[j]-bin_width/2 and r< rs[j]+bin_width/2): k=j totA_ep100000[k] +=1 if (target_times[index][n]>0.0): sucA_ep100000[k] +=1 for n in range(N2): index = 45 #### episode 500000 r = initial_radii2[index][n] scalar = np.cos(initial_directions2[index][n]) k=-1 for j in range(len(rs)): if (r>= rs[j]-bin_width/2 and r< rs[j]+bin_width/2): k=j totB_ep100000[k] +=1 if (target_times2[index][n]>0.0): sucB_ep100000[k] +=1 #################################################################### #################################################################### #################################################################### xfig,yfig=7.0,8.5 factor = xfig/yfig ygap = 0.05 dy = 0.18 x1,y1 = 0.08,4*ygap + 3*dy dx1,dy1=0.4,dy x2,y2 = 0.58,y1 dx2,dy2=dx1,dy1 x3,y3 =x1, 3*ygap + 2*dy dx3,dy3=dx1,dy1 x4,y4 =x2, y3 dx4,dy4=dx1,dy1 x5,y5 =x1, 2*ygap + dy dx5,dy5=dx1,dy1 x6,y6 =x2, y5 dx6,dy6=dx1,dy1 x7,y7 =x1, ygap dx7,dy7=dx1,dy1 x8,y8 =x2, y7 dx8,dy8=dx1,dy1 x9,y9 = x1 + dx1/8, y1 + dy1/3.5 dx9,dy9 = dx1/3,dy1/2 x10,y10 = x2 + dx1/8, y9 dx10,dy10 = dx9,dy9 with PdfPages('fig4.pdf') as pdf: fig = plt.figure(figsize=(xfig,yfig)) plt.rc('text', usetex=True) plt.rc('text.latex', preamble = ','.join('''\usepackage{txfonts} \usepackage{lmodern}'''.split())) ######################## panel A ######################################## panel = fig.add_axes([x1, y1, dx1, dy1]) panel.tick_params(axis='both',which='both',direction='in',bottom=True,top=True,left=True,right=True) for tick in panel.xaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsize) for tick in panel.yaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsize) panel.set_xlabel(r'episodes',fontsize=axislabelfontsize) panel.set_ylabel(r'fraction of successful agents',fontsize=axislabelfontsize) # ~ panel.set_xlim(0.8,280000) panel.set_xlim(0.8,140000) panel.set_ylim(0,1.0) panel.set_xscale('log') panel.yaxis.set_major_locator(MultipleLocator(0.1)) panel.yaxis.set_minor_locator(MultipleLocator(0.05)) panel.xaxis.set_major_locator(LogLocator(base=10.0, subs=[1.0], numticks=10)) # Major ticks at 10^x panel.xaxis.set_minor_locator(LogLocator(base=10.0, subs=np.arange(1, 10) * 0.1, numticks=100)) # Minor ticks at 1-9 panel.xaxis.set_major_formatter(LogFormatterMathtext()) # Forces 10^x notation panel.xaxis.set_minor_formatter(NullFormatter()) # plot avg line dummyX = [1,10] dummyY = [2,3] panel.plot(dummyX,dummyY,'o-',color=color3,markersize=markersize+3,linewidth=linewidth2,zorder=3,label=r'$r_{\rm ini}=2\sigma$') panel.plot(dummyX,dummyY,'o-',color=color2,markersize=markersize+3,linewidth=linewidth2,zorder=3,label=r'$r_{\rm ini}=5\sigma$') panel.plot(dummyX,dummyY,'o-',color=color1,markersize=markersize+3,linewidth=linewidth2,zorder=3,label=r'$r_{\rm ini}=8\sigma$') # ~ panel.plot(dummyX,dummyY,'-.',color='m',markersize=2,linewidth=linewidth3,zorder=5,label=r'weight avg.') panel.plot(dummyX,dummyY,'s--',color='darkgray',markersize=markersize+4,linewidth=linewidth,zorder=4,label=r'$\sigma/2 < r_{\rm ini} \leq \tilde{R}$') panel.plot(dummyX,dummyY,'--',color='k',markersize=2,linewidth=linewidth,zorder=5,label=r'init. policy') panel.plot(episodes_rini[index2],n_targets_rini[index2]*1.0/N,'-',color=color3,markersize=2,linewidth=linewidth2,zorder=3) panel.plot(episodes_rini[index5],n_targets_rini[index5]*1.0/N,'-',color=color2,markersize=2,linewidth=linewidth2,zorder=3) panel.plot(episodes_rini[index8],n_targets_rini[index8]*1.0/N,'-',color=color1,markersize=2,linewidth=linewidth2,zorder=3) # ~ panel.plot(episodes,n_targets_weightavg*1.0/N,'-.',color='m',markersize=2,linewidth=linewidth3,zorder=5) panel.plot(episodes,n_targets*1.0/N,'--',color='darkgray',markersize=2,linewidth=linewidth,zorder=4) panel.legend(loc='upper left', bbox_to_anchor=(0.6, 1.46),ncol=2,fontsize=legendfontsize-1,handlelength=3.5,labelspacing=0.2,framealpha=1.0) panel.text(0.03,0.9,r'{\bf (a) Type A}',fontsize=legendfontsize,transform=panel.transAxes) ########## panel = fig.add_axes([x9, y9, dx9, dy9]) panel.tick_params(axis='both',which='both',direction='in',bottom=True,top=True,left=True,right=True) for tick in panel.xaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsizeinset) for tick in panel.yaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsizeinset) panel.set_xlabel(r'$r_{\rm ini}/\sigma$',fontsize=axislabelfontsizeinset,labelpad=-0.6) panel.set_xlim(0,10.5) panel.set_ylim(0,1.0) panel.xaxis.set_major_locator(MultipleLocator(2)) panel.xaxis.set_minor_locator(MultipleLocator(1)) panel.yaxis.set_major_locator(MultipleLocator(0.2)) panel.yaxis.set_minor_locator(MultipleLocator(0.1)) # plot avg line panel.plot(rinis_fixed,final_n_targets_rini*1.0/N,'r-',markersize=2,linewidth=linewidth,zorder=3) panel.plot(rs,sucA_ep100000/totA_ep100000,'--', color='darkgray') panel.text(0.1,0.8,r'$10^5$-th episode',fontsize=legendfontsize-1,transform=panel.transAxes) ######################## panel B ######################################## panel = fig.add_axes([x2, y2, dx2, dy2]) panel.tick_params(axis='both',which='both',direction='in',bottom=True,top=True,left=True,right=True) for tick in panel.xaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsize) for tick in panel.yaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsize) panel.set_xlabel(r'episodes',fontsize=axislabelfontsize) panel.set_ylabel(r'fraction of successful agents',fontsize=axislabelfontsize) # ~ panel.set_xlim(0.8,280000) panel.set_xlim(0.8,140000) panel.set_ylim(0.0,1.0) panel.set_xscale('log') panel.yaxis.set_major_locator(MultipleLocator(0.1)) panel.yaxis.set_minor_locator(MultipleLocator(0.05)) panel.xaxis.set_major_locator(LogLocator(base=10.0, subs=[1.0], numticks=10)) # Major ticks at 10^x panel.xaxis.set_minor_locator(LogLocator(base=10.0, subs=np.arange(1, 10) * 0.1, numticks=100)) # Minor ticks at 1-9 panel.xaxis.set_major_formatter(LogFormatterMathtext()) # Forces 10^x notation panel.xaxis.set_minor_formatter(NullFormatter()) # plot avg line panel.plot(episodes2_rini[index2],n_targets2_rini[index2]*1.0/N2,'-',color=color3,markersize=2,linewidth=linewidth2,zorder=3,label=r'$r_{\rm ini}=2\sigma$') panel.plot(episodes2_rini[index5],n_targets2_rini[index5]*1.0/N2,'-',color=color2,markersize=2,linewidth=linewidth2,zorder=3,label=r'$r_{\rm ini}=5\sigma$') panel.plot(episodes2_rini[index8],n_targets2_rini[index8]*1.0/N2,'-',color=color1,markersize=2,linewidth=linewidth2,zorder=3,label=r'$r_{\rm ini}=8\sigma$') # ~ panel.plot(episodes2,n_targets_weightavg2*1.0/N,'-.',color='m',markersize=2,linewidth=linewidth3,zorder=5,label=r'weight avg.') panel.plot(episodes2,n_targets2*1.0/N2,'--',color='darkgray',markersize=2,linewidth=linewidth,zorder=4,label=r'$\sigma/2 < r_{\rm ini} \leq \tilde{R}$') #~ panel.legend(loc='upper left', bbox_to_anchor=(0.01, 0.87),ncol=1,fontsize=legendfontsize-1,handlelength=2.5,labelspacing=0.2,framealpha=1.0) panel.text(0.03,0.9,r'{\bf (b) Type C}',fontsize=legendfontsize,transform=panel.transAxes) ########## panel = fig.add_axes([x10, y10, dx10, dy10]) panel.tick_params(axis='both',which='both',direction='in',bottom=True,top=True,left=True,right=True) for tick in panel.xaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsizeinset) for tick in panel.yaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsizeinset) panel.set_xlabel(r'$r_{\rm ini}/\sigma$',fontsize=axislabelfontsizeinset,labelpad=-0.6) panel.set_xlim(0,10.5) panel.set_ylim(0.7,1.1) panel.xaxis.set_major_locator(MultipleLocator(2)) panel.xaxis.set_minor_locator(MultipleLocator(1)) panel.yaxis.set_major_locator(MultipleLocator(0.2)) panel.yaxis.set_minor_locator(MultipleLocator(0.1)) # plot avg line panel.plot(rinis_fixed,final_n_targets2_rini*1.0/N,'r-',markersize=2,linewidth=linewidth,zorder=3) panel.plot(rs,sucB_ep100000/totB_ep100000,'--', color='darkgray') panel.text(0.1,0.8,r'$10^5$-th episode',fontsize=legendfontsize-1,transform=panel.transAxes) ######################## panel C ######################################## panel = fig.add_axes([x3, y3, dx3, dy3]) panel.tick_params(axis='both',which='both',direction='in',bottom=True,top=True,left=True,right=True) for tick in panel.xaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsize) for tick in panel.yaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsize) panel.set_xlabel(r'$r/\sigma$',fontsize=axislabelfontsize) panel.set_ylabel(r'switch. prob. $(\phi \!=\! 0)$',fontsize=axislabelfontsize) panel.set_xlim(-0.2,10.7) panel.xaxis.set_major_locator(MultipleLocator(2)) panel.xaxis.set_minor_locator(MultipleLocator(1)) panel.set_ylim(0.0004,0.8) panel.set_yscale('log') panel.plot(rinis[1:M-1],dataTypeA_ep100000_rini[index2][0][1:M-1],'o',color=color3,markersize=markersize,linewidth=linewidth,zorder=3) panel.plot(rinis[0],dataTypeA_ep100000_rini[index2][0][0],'o',color=color3,markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[M-1],dataTypeA_ep100000_rini[index2][0][M-1],'o',color=color3,markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[1:M-1],dataTypeA_ep100000_rini[index5][0][1:M-1],'o',color=color2,markersize=markersize,linewidth=linewidth,zorder=3) panel.plot(rinis[0],dataTypeA_ep100000_rini[index5][0][0],'o',color=color2,markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[M-1],dataTypeA_ep100000_rini[index5][0][M-1],'o',color=color2,markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[1:M-1],dataTypeA_ep100000_rini[index8][0][1:M-1],'o',color=color1,markersize=markersize,linewidth=linewidth,zorder=3) panel.plot(rinis[0],dataTypeA_ep100000_rini[index8][0][0],'o',color=color1,markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[M-1],dataTypeA_ep100000_rini[index8][0][M-1],'o',color=color1,markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[1:M-1],dataTypeA_ep100000[0][1:M-1],'s',color='darkgray',markersize=markersize-0.4,linewidth=linewidth,zorder=3) panel.plot(rinis[0],dataTypeA_ep100000[0][0],'s',color='darkgray',markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[M-1],dataTypeA_ep100000[0][M-1],'s',color='darkgray',markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis,pBP,'k--',linewidth=linewidth,zorder=3) panel.axvline(x=0.5, ymin=0.0, ymax=0.75, color='gray', linestyle='-', linewidth=1) panel.text(0.08,0.74,r'$\sigma/2$',fontsize=axisticslabelfontsize,color='gray',transform=panel.transAxes) panel.axvline(x=10.0, ymin=0.0, ymax=0.75, color='gray', linestyle='-', linewidth=1) panel.text(0.87,0.74,r'$\tilde{R}$',fontsize=axisticslabelfontsize,color='gray',transform=panel.transAxes) panel.axvline(x=2, linestyle='--', linewidth=0.5, color=color3) panel.axvline(x=5, linestyle='--', linewidth=0.5, color=color2) panel.axvline(x=8, linestyle='--', linewidth=0.5, color=color1) panel.text(0.03,0.9,r'{\bf (c) Type A}',fontsize=legendfontsize,transform=panel.transAxes) ######################## panel D ######################################## panel = fig.add_axes([x4, y4, dx4, dy4]) panel.tick_params(axis='both',which='both',direction='in',bottom=True,top=True,left=True,right=True) for tick in panel.xaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsize) for tick in panel.yaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsize) panel.set_xlabel(r'$r/\sigma$',fontsize=axislabelfontsize) panel.set_ylabel(r'switch. prob. $(\phi \!=\! 1)$',fontsize=axislabelfontsize) panel.set_xlim(-0.2,10.7) panel.xaxis.set_major_locator(MultipleLocator(2)) panel.xaxis.set_minor_locator(MultipleLocator(1)) panel.set_ylim(0.00003,0.8) panel.set_yscale('log') panel.plot(rinis[1:M-1],dataTypeA_ep100000_rini[index2][1][1:M-1],'o',color=color3,markersize=markersize,linewidth=linewidth,zorder=3) panel.plot(rinis[0],dataTypeA_ep100000_rini[index2][1][0],'o',color=color3,markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[M-1],dataTypeA_ep100000_rini[index2][1][M-1],'o',color=color3,markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[1:M-1],dataTypeA_ep100000_rini[index5][1][1:M-1],'o',color=color2,markersize=markersize,linewidth=linewidth,zorder=3) panel.plot(rinis[0],dataTypeA_ep100000_rini[index5][1][0],'o',color=color2,markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[M-1],dataTypeA_ep100000_rini[index5][1][M-1],'o',color=color2,markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[1:M-1],dataTypeA_ep100000_rini[index8][1][1:M-1],'o',color=color1,markersize=markersize,linewidth=linewidth,zorder=3) panel.plot(rinis[0],dataTypeA_ep100000_rini[index8][1][0],'o',color=color1,markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[M-1],dataTypeA_ep100000_rini[index8][1][M-1],'o',color=color1,markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[1:M-1],dataTypeA_ep100000[1][1:M-1],'s',color='darkgray',markersize=markersize-0.4,linewidth=linewidth,zorder=3) panel.plot(rinis[0],dataTypeA_ep100000[1][0],'s',color='darkgray',markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[M-1],dataTypeA_ep100000[1][M-1],'s',color='darkgray',markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis,pABP,'k--',linewidth=linewidth,zorder=3) panel.axvline(x=0.5, ymin=0.0, ymax=0.75, color='gray', linestyle='-', linewidth=1) panel.text(0.08,0.74,r'$\sigma/2$',fontsize=axisticslabelfontsize,color='gray',transform=panel.transAxes) panel.axvline(x=10.0, ymin=0.0, ymax=0.75, color='gray', linestyle='-', linewidth=1) panel.text(0.87,0.74,r'$\tilde{R}$',fontsize=axisticslabelfontsize,color='gray',transform=panel.transAxes) panel.axvline(x=2, linestyle='--', linewidth=0.5, color=color3) panel.axvline(x=5, linestyle='--', linewidth=0.5, color=color2) panel.axvline(x=8, linestyle='--', linewidth=0.5, color=color1) panel.text(0.03,0.9,r'{\bf (d) Type A}',fontsize=legendfontsize,transform=panel.transAxes) ######################## panel E ######################################## panel = fig.add_axes([x5, y5, dx5, dy5]) panel.tick_params(axis='both',which='both',direction='in',bottom=True,top=True,left=True,right=True) for tick in panel.xaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsize) for tick in panel.yaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsize) panel.set_xlabel(r'$r/\sigma$',fontsize=axislabelfontsize) panel.set_ylabel(r'switch. prob. $(\phi \!=\! 0, \, \omega \!=\! 0)$',fontsize=axislabelfontsize) panel.set_xlim(-0.2,10.7) panel.xaxis.set_major_locator(MultipleLocator(2)) panel.xaxis.set_minor_locator(MultipleLocator(1)) panel.set_ylim(0.0008,0.8) panel.set_yscale('log') panel.plot(rinis[1:M-1],dataTypeC_ep100000_rini[index2][0][1:M-1],'o',color=color3,markersize=markersize,linewidth=linewidth,zorder=3) panel.plot(rinis[0],dataTypeC_ep100000_rini[index2][0][0],'o',color=color3,markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[M-1],dataTypeC_ep100000_rini[index2][0][M-1],'o',color=color3,markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[1:M-1],dataTypeC_ep100000_rini[index5][0][1:M-1],'o',color=color2,markersize=markersize,linewidth=linewidth,zorder=3) panel.plot(rinis[0],dataTypeC_ep100000_rini[index5][0][0],'o',color=color2,markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[M-1],dataTypeC_ep100000_rini[index5][0][M-1],'o',color=color2,markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[1:M-1],dataTypeC_ep100000_rini[index8][0][1:M-1],'o',color=color1,markersize=markersize,linewidth=linewidth,zorder=3) panel.plot(rinis[0],dataTypeC_ep100000_rini[index8][0][0],'o',color=color1,markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[M-1],dataTypeC_ep100000_rini[index8][0][M-1],'o',color=color1,markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[1:M-1],dataTypeC_ep100000[0][1:M-1],'s',color='darkgray',markersize=markersize-0.4,linewidth=linewidth,zorder=3) panel.plot(rinis[0],dataTypeC_ep100000[0][0],'s',color='darkgray',markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[M-1],dataTypeC_ep100000[0][M-1],'s',color='darkgray',markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis,pBP,'k--',linewidth=linewidth,zorder=3) panel.axvline(x=0.5, ymin=0.0, ymax=0.75, color='gray', linestyle='-', linewidth=1) panel.text(0.08,0.74,r'$\sigma/2$',fontsize=axisticslabelfontsize,color='gray',transform=panel.transAxes) panel.axvline(x=10.0, ymin=0.0, ymax=0.75, color='gray', linestyle='-', linewidth=1) panel.text(0.87,0.74,r'$\tilde{R}$',fontsize=axisticslabelfontsize,color='gray',transform=panel.transAxes) panel.axvline(x=2, linestyle='--', linewidth=0.5, color=color3) panel.axvline(x=5, linestyle='--', linewidth=0.5, color=color2) panel.axvline(x=8, linestyle='--', linewidth=0.5, color=color1) panel.text(0.03,0.9,r'{\bf (e) Type C}',fontsize=legendfontsize,transform=panel.transAxes) ######################## panel F ######################################## panel = fig.add_axes([x6, y6, dx6, dy6]) panel.tick_params(axis='both',which='both',direction='in',bottom=True,top=True,left=True,right=True) for tick in panel.xaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsize) for tick in panel.yaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsize) panel.set_xlabel(r'$r/\sigma$',fontsize=axislabelfontsize) panel.set_ylabel(r'switch. prob. $(\phi \!=\! 1, \, \omega \!=\! 0)$',fontsize=axislabelfontsize) panel.set_xlim(-0.2,10.7) panel.xaxis.set_major_locator(MultipleLocator(2)) panel.xaxis.set_minor_locator(MultipleLocator(1)) panel.set_ylim(0.0005,0.8) panel.set_yscale('log') panel.plot(rinis[1:M-1],dataTypeC_ep100000_rini[index2][1][1:M-1],'o',color=color3,markersize=markersize,linewidth=linewidth,zorder=3) panel.plot(rinis[0],dataTypeC_ep100000_rini[index2][1][0],'o',color=color3,markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[M-1],dataTypeC_ep100000_rini[index2][1][M-1],'o',color=color3,markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[1:M-1],dataTypeC_ep100000_rini[index5][1][1:M-1],'o',color=color2,markersize=markersize,linewidth=linewidth,zorder=3) panel.plot(rinis[0],dataTypeC_ep100000_rini[index5][1][0],'o',color=color2,markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[M-1],dataTypeC_ep100000_rini[index5][1][M-1],'o',color=color2,markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[1:M-1],dataTypeC_ep100000_rini[index8][1][1:M-1],'o',color=color1,markersize=markersize,linewidth=linewidth,zorder=3) panel.plot(rinis[0],dataTypeC_ep100000_rini[index8][1][0],'o',color=color1,markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[M-1],dataTypeC_ep100000_rini[index8][1][M-1],'o',color=color1,markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[1:M-1],dataTypeC_ep100000[1][1:M-1],'s',color='darkgray',markersize=markersize-0.4,linewidth=linewidth,zorder=3) panel.plot(rinis[0],dataTypeC_ep100000[1][0],'s',color='darkgray',markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[M-1],dataTypeC_ep100000[1][M-1],'s',color='darkgray',markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis,pABP,'k--',linewidth=linewidth,zorder=3) panel.axvline(x=0.5, ymin=0.0, ymax=0.75, color='gray', linestyle='-', linewidth=1) panel.text(0.08,0.74,r'$\sigma/2$',fontsize=axisticslabelfontsize,color='gray',transform=panel.transAxes) panel.axvline(x=10.0, ymin=0.0, ymax=0.75, color='gray', linestyle='-', linewidth=1) panel.text(0.87,0.74,r'$\tilde{R}$',fontsize=axisticslabelfontsize,color='gray',transform=panel.transAxes) panel.axvline(x=2, linestyle='--', linewidth=0.5, color=color3) panel.axvline(x=5, linestyle='--', linewidth=0.5, color=color2) panel.axvline(x=8, linestyle='--', linewidth=0.5, color=color1) panel.text(0.03,0.9,r'{\bf (f) Type C}',fontsize=legendfontsize,transform=panel.transAxes) ######################## panel G ######################################## panel = fig.add_axes([x7, y7, dx7, dy7]) panel.tick_params(axis='both',which='both',direction='in',bottom=True,top=True,left=True,right=True) for tick in panel.xaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsize) for tick in panel.yaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsize) panel.set_xlabel(r'$r/\sigma$',fontsize=axislabelfontsize) panel.set_ylabel(r'switch. prob. $(\phi \!=\! 0, \, \omega \!=\! 1)$',fontsize=axislabelfontsize) panel.set_xlim(-0.2,10.7) panel.xaxis.set_major_locator(MultipleLocator(2)) panel.xaxis.set_minor_locator(MultipleLocator(1)) panel.set_ylim(0.002,0.8) panel.set_yscale('log') panel.plot(rinis[1:M-1],dataTypeC_ep100000_rini[index2][2][1:M-1],'o',color=color3,markersize=markersize,linewidth=linewidth,zorder=3) panel.plot(rinis[0],dataTypeC_ep100000_rini[index2][2][0],'o',color=color3,markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[M-1],dataTypeC_ep100000_rini[index2][2][M-1],'o',color=color3,markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[1:M-1],dataTypeC_ep100000_rini[index5][2][1:M-1],'o',color=color2,markersize=markersize,linewidth=linewidth,zorder=3) panel.plot(rinis[0],dataTypeC_ep100000_rini[index5][2][0],'o',color=color2,markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[M-1],dataTypeC_ep100000_rini[index5][2][M-1],'o',color=color2,markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[1:M-1],dataTypeC_ep100000_rini[index8][2][1:M-1],'o',color=color1,markersize=markersize,linewidth=linewidth,zorder=3) panel.plot(rinis[0],dataTypeC_ep100000_rini[index8][2][0],'o',color=color1,markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[M-1],dataTypeC_ep100000_rini[index8][2][M-1],'o',color=color1,markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[1:M-1],dataTypeC_ep100000[2][1:M-1],'s',color='darkgray',markersize=markersize-0.4,linewidth=linewidth,zorder=3) panel.plot(rinis[0],dataTypeC_ep100000[2][0],'s',color='darkgray',markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[M-1],dataTypeC_ep100000[2][M-1],'s',color='darkgray',markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis,pBP,'k--',linewidth=linewidth,zorder=3) panel.axvline(x=0.5, ymin=0.0, ymax=0.75, color='gray', linestyle='-', linewidth=1) panel.text(0.08,0.74,r'$\sigma/2$',fontsize=axisticslabelfontsize,color='gray',transform=panel.transAxes) panel.axvline(x=10.0, ymin=0.0, ymax=0.75, color='gray', linestyle='-', linewidth=1) panel.text(0.87,0.74,r'$\tilde{R}$',fontsize=axisticslabelfontsize,color='gray',transform=panel.transAxes) panel.axvline(x=2, linestyle='--', linewidth=0.5, color=color3) panel.axvline(x=5, linestyle='--', linewidth=0.5, color=color2) panel.axvline(x=8, linestyle='--', linewidth=0.5, color=color1) panel.text(0.03,0.9,r'{\bf (g) Type C}',fontsize=legendfontsize,transform=panel.transAxes) ######################## panel H ######################################## panel = fig.add_axes([x8, y8, dx8, dy8]) panel.tick_params(axis='both',which='both',direction='in',bottom=True,top=True,left=True,right=True) for tick in panel.xaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsize) for tick in panel.yaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsize) panel.set_xlabel(r'$r/\sigma$',fontsize=axislabelfontsize) panel.set_ylabel(r'switch. prob. $(\phi \!=\! 1, \, \omega \!=\! 1)$',fontsize=axislabelfontsize) panel.set_xlim(-0.2,10.7) panel.xaxis.set_major_locator(MultipleLocator(2)) panel.xaxis.set_minor_locator(MultipleLocator(1)) panel.set_ylim(0.00003,0.2) panel.set_yscale('log') panel.plot(rinis[1:M-1],dataTypeC_ep100000_rini[index2][3][1:M-1],'o',color=color3,markersize=markersize,linewidth=linewidth,zorder=3) panel.plot(rinis[0],dataTypeC_ep100000_rini[index2][3][0],'o',color=color3,markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[M-1],dataTypeC_ep100000_rini[index2][3][M-1],'o',color=color3,markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[1:M-1],dataTypeC_ep100000_rini[index5][3][1:M-1],'o',color=color2,markersize=markersize,linewidth=linewidth,zorder=3) panel.plot(rinis[0],dataTypeC_ep100000_rini[index5][3][0],'o',color=color2,markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[M-1],dataTypeC_ep100000_rini[index5][3][M-1],'o',color=color2,markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[1:M-1],dataTypeC_ep100000_rini[index8][3][1:M-1],'o',color=color1,markersize=markersize,linewidth=linewidth,zorder=3) panel.plot(rinis[0],dataTypeC_ep100000_rini[index8][3][0],'o',color=color1,markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[M-1],dataTypeC_ep100000_rini[index8][3][M-1],'o',color=color1,markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[1:M-1],dataTypeC_ep100000[3][1:M-1],'s',color='darkgray',markersize=markersize-0.4,linewidth=linewidth,zorder=3) panel.plot(rinis[0],dataTypeC_ep100000[3][0],'s',color='darkgray',markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis[M-1],dataTypeC_ep100000[3][M-1],'s',color='darkgray',markersize=markersize+2,linewidth=linewidth,zorder=3) panel.plot(rinis,pABP,'k--',linewidth=linewidth,zorder=3) panel.axvline(x=0.5, ymin=0.0, ymax=0.75, color='gray', linestyle='-', linewidth=1) panel.text(0.08,0.74,r'$\sigma/2$',fontsize=axisticslabelfontsize,color='gray',transform=panel.transAxes) panel.axvline(x=10.0, ymin=0.0, ymax=0.75, color='gray', linestyle='-', linewidth=1) panel.text(0.87,0.74,r'$\tilde{R}$',fontsize=axisticslabelfontsize,color='gray',transform=panel.transAxes) panel.axvline(x=2, linestyle='--', linewidth=0.5, color=color3) panel.axvline(x=5, linestyle='--', linewidth=0.5, color=color2) panel.axvline(x=8, linestyle='--', linewidth=0.5, color=color1) panel.text(0.03,0.9,r'{\bf (h) Type C}',fontsize=legendfontsize,transform=panel.transAxes) pdf.savefig(fig) return def global_variables (): # variabili comuni utili per le varie figure global axisticslabelfontsize global axisticslabelfontsizeinset global axislabelfontsize global axislabelfontsizeinset global legendfontsize global linewidth global linewidth2 global linewidth3 global linewidth4 global markersize # colors (colorblind safe) global color1 global color2 global color3 axisticslabelfontsize=9 axisticslabelfontsizeinset=7 axislabelfontsize=11 axislabelfontsizeinset=8 legendfontsize=10 linewidth = 2.0 linewidth2 = 3.0 linewidth3 = 1.5 linewidth4 = 6.0 markersize = 1.0 color1 = "#d95f02" color2 = "#1b9e77" color3 = "#7570b3" return def main(): global_variables () Dr = 0.01 R_target = 0.5 R_max = 10.0 M = 2 + int((R_max-R_target)/Dr) # number of distance bins print (M) rinis = np.zeros((M)) for i in range(M): rinis[i] = R_target + 0.5*((i-1)*Dr + (i)*Dr) print (rinis) ####### Type A N_bunches = 40 N_runs_bunch = 250 # number of completely independent runs N_episodes = 1000000 # number of episodes. Each episode lasts for a time_single_episode foldername = '../../../../TargetSearching_Local/ReinforcementLearning/Paper4/RESULTS/TypeA_plain_Pe100/' # ~ #~ foldername = '../../../../../../../../media/mik/Elements/Physics/Innsbruck/TargetSearching_Local/ReinforcementLearning/Paper4/RESULTS/TypeA_plain_Pe100/' ####### Type C N_bunches2 = 40 N_runs_bunch2 = 250 # number of completely independent runs N_episodes2 = 1000000 # number of episodes. Each episode lasts for a time_single_episode foldername2 = '../../../../TargetSearching_Local/ReinforcementLearning/Paper4/RESULTS/TypeC_plain_Pe100/' # ~ #~ foldername2 = '../../../../../../../../media/mik/Elements/Physics/Innsbruck/TargetSearching_Local/ReinforcementLearning/Paper4/RESULTS/TypeC_plain_Pe100/' ####### Type B N_bunches3 = 40 N_runs_bunch3 = 250 # number of completely independent runs N_episodes3 = 1000000 # number of episodes. Each episode lasts for a time_single_episode foldername3 = '../../../../TargetSearching_Local/ReinforcementLearning/Paper4/RESULTS/TypeD_plain_Pe100/' # ~ #~ foldername3 = '../../../../../../../../media/mik/Elements/Physics/Innsbruck/TargetSearching_Local/ReinforcementLearning/Paper4/RESULTS/TypeB_plain_Pe100/' #### test xs,ys,zs=read_TargetTimes_given_policy(foldername2,'LearnedPolicy_Modified') testvalue = 0.0 sumvalue=0.0 test2value=0.0 for x in xs: sumvalue+=1.0 if (x>0.00001): testvalue+=1.0 test2value+=x test2value = test2value/testvalue testvalue = testvalue/sumvalue # ~ # plots make_figure1(N_bunches,N_runs_bunch,N_episodes,foldername,N_bunches2,N_runs_bunch2,N_episodes2,foldername2,N_bunches3,N_runs_bunch3,N_episodes3,foldername3,testvalue,test2value) make_figure2(M,rinis, N_bunches,N_runs_bunch,N_episodes,foldername,N_bunches2,N_runs_bunch2,N_episodes2,foldername2,N_bunches3,N_runs_bunch3,N_episodes3,foldername3) # ~ ####### Type A radius 5 # ~ N_bunches = 40 # ~ N_runs_bunch = 250 # number of completely independent runs # ~ N_episodes = 100000 # number of episodes. Each episode lasts for a time_single_episode # ~ foldername = '../../../../TargetSearching_Local/ReinforcementLearning/Paper4/RESULTS/' # ~ #~ foldername = '../../../../../../../../media/mik/Elements/Physics/Innsbruck/TargetSearching_Local/ReinforcementLearning/Paper4/RESULTS/' # ~ ####### Type C radius 5 # ~ N_bunches2 = 40 # ~ N_runs_bunch2 = 250 # number of completely independent runs # ~ N_episodes2 = 100000 # number of episodes. Each episode lasts for a time_single_episode # ~ foldername2 = '../../../../TargetSearching_Local/ReinforcementLearning/Paper4/RESULTS/' # ~ #~ foldername2 = '../../../../../../../../media/mik/Elements/Physics/Innsbruck/TargetSearching_Local/ReinforcementLearning/Paper4/RESULTS/' # ~ make_figure3(M,rinis,N_bunches,N_runs_bunch,N_episodes,foldername,N_bunches2,N_runs_bunch2,N_episodes2,foldername2) print (testvalue,test2value) main()