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 Aux import * def func(x,b): return np.exp(-(x**b)) def make_figure (episodes,avg_time,avg_time_passive,avg_time_optimal,times,yfunc,Pes,TRTs,times1,times2,times5,times10,times20,times50,times100,times200,times500,times1000,times2000): # create benchmark curves benchmark_passive = np.zeros((len(episodes))) benchmark_optimal = np.zeros((len(episodes))) for i in range(len(episodes)): benchmark_passive[i] = avg_time_passive benchmark_optimal[i] = avg_time_optimal benchmark_passive_Pes = np.zeros((len(Pes))) benchmark_optimal_Pes = np.zeros((len(Pes))) for i in range(len(Pes)): benchmark_passive_Pes[i] = avg_time_passive benchmark_optimal_Pes[i] = avg_time_optimal # setto alcune variabili comuni axisticslabelfontsize=9 axisticslabelfontsizeinset=7 axislabelfontsize=11 axislabelfontsizeinset=9 xfig,yfig=7.0,2.5 factor = xfig/yfig x1,y1 = 0.13,0.16 #~ x1,y1 = 0.12,0.16 # to uncomment if we wnat to add the fit line dx1,dy1=0.5,0.82 x2,y2 = x1+dx1,y1 dx2,dy2 = 0.28,dy1 xlim_left,xlim_middle,xlim_right = -1,50,1000 ylim_low,ylim_up = 0.0, 2.92 x3,y3 = x1+dx1+0.05,0.48 dx3,dy3 = 0.21,0.44 with PdfPages('fig1.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 0 ################################## to uncomment if we wnat to add the fit line #~ try: #~ panel = fig.add_axes([0, 0, 1, 1]) #~ panel.spines['right'].set_visible(False) #~ panel.spines['top'].set_visible(False) #~ panel.spines['bottom'].set_visible(False) #~ panel.spines['left'].set_visible(False) #~ panel.axes.get_xaxis().set_ticks([]) #~ panel.axes.get_yaxis().set_ticks([]) #~ Deltay1 = 0.075 #~ Deltay2 = 0.672 #~ l1 = lines.Line2D([x2+dx2+0.02, x2+dx2+0.02], [y2, y2+Deltay1], transform=panel.transAxes, color='black', alpha=1.0, linewidth=0.5) #~ panel.add_line(l1) #~ l1 = lines.Line2D([x2+dx2+0.02, x2+dx2+0.02], [y2+Deltay1,y2+Deltay2], transform=panel.transAxes, color='black', alpha=1.0, linewidth=0.5) #~ panel.add_line(l1) #~ l1 = lines.Line2D([x2+dx2+0.012, x2+dx2+0.02], [y2+Deltay1,y2+Deltay1], transform=panel.transAxes, color='black', alpha=1.0, linewidth=0.5) #~ panel.add_line(l1) #~ l1 = lines.Line2D([x2+dx2+0.012, x2+dx2+0.02], [y2,y2], transform=panel.transAxes, color='black', alpha=1.0, linewidth=0.5) #~ panel.add_line(l1) #~ l1 = lines.Line2D([x2+dx2+0.012, x2+dx2+0.02], [y2+Deltay2,y2+Deltay2], transform=panel.transAxes, color='black', alpha=1.0, linewidth=0.5) #~ panel.add_line(l1) #~ panel.text(x2+dx2+0.03,y2+Deltay1/4,r'C',fontsize=axislabelfontsize+1,transform=panel.transAxes) #~ panel.text(x2+dx2+0.03,y2+Deltay2/2,r'D',fontsize=axislabelfontsize+1,transform=panel.transAxes) #~ except: #~ print ('error in panel 0') ######################## 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=False) 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.xaxis.set_label_coords(.82, -.1) panel.set_ylabel(r'avg. time to reach the target [$\tau$]',fontsize=axislabelfontsize) panel.set_xlim(xlim_left,xlim_middle) panel.set_ylim(ylim_low,ylim_up) panel.xaxis.set_major_locator(MultipleLocator(10)) panel.xaxis.set_minor_locator(MultipleLocator(2)) panel.yaxis.set_major_locator(MultipleLocator(0.5)) panel.yaxis.set_minor_locator(MultipleLocator(0.1)) t = panel.text(0.1,0.88,r'Pe = $100$',fontsize=axislabelfontsize+1,transform=panel.transAxes) t.set_bbox(dict(facecolor='white', alpha=0.8, edgecolor='none',pad=1.5)) # plot avg line panel.plot(episodes,avg_time,'-',color='darkgreen',markersize=2,linewidth=2,zorder=3,label=r'learning agent (avg.)') panel.plot(episodes,benchmark_passive,'b--',markersize=2,linewidth=1.5,zorder=3,label=r'passive particle (avg.)') #~ panel.plot(episodes,benchmark_optimal,'k--',markersize=2,linewidth=1.5,zorder=3,label=r'optimal agent (avg.)') #~ panel.plot(episodes,yfunc,'r--',markersize=2,linewidth=0.9,zorder=3,label=r'$C + D \exp(-x^{\beta}) \; , \quad \beta=0.245\pm0.001$') # to uncomment if we wnat to add the fit line # plot boxes c = "green" xwidth = 1.5 binwidth = 2 for index in range(len(episodes)): if (index%binwidth == 0): x = episodes[index] data = np.zeros((len(times[index]))) for k in range(len(times[index])): data[k]=times[index][k] panel.boxplot(data, sym='', whis=[10,90], positions=[x], widths = xwidth, manage_xticks=False, autorange=False, patch_artist=True,boxprops=dict(facecolor='lightgreen', color=c),capprops=dict(color=c),whiskerprops=dict(color=c),medianprops=dict(color=c)) panel.legend(loc='upper right', bbox_to_anchor=(0.99, 0.8),ncol=1,fontsize=8,handlelength=1.5,labelspacing=0.2,framealpha=1.0) factorx = 1.2 # = 1 for 40 episodes xr,yr=34,2.4 dxr,dyr=5.8*factorx,0.17 epsxr,epsyr = 1*factorx,0.22 rect = patches.Rectangle((xr,yr),dxr,dyr, facecolor='lightgreen',edgecolor='g') panel.add_patch(rect) line = lines.Line2D([xr, xr-dxr/2], [yr+dyr/2, yr+dyr/2], color='g', linewidth=0.8) panel.add_line(line) line = lines.Line2D([xr-dxr/2,xr-dxr/2], [yr+dyr/3, yr+2*dyr/3], color='g', linewidth=0.8) panel.add_line(line) line = lines.Line2D([xr+dxr, xr+dxr+dxr/2], [yr+dyr/2, yr+dyr/2], color='g', linewidth=0.8) panel.add_line(line) line = lines.Line2D([xr+dxr+dxr/2,xr+dxr+dxr/2], [yr+dyr/3, yr+2*dyr/3], color='g', linewidth=0.8) panel.add_line(line) line = lines.Line2D([xr+dxr/2,xr+dxr/2], [yr, yr+dyr], color='g', linewidth=0.8) panel.add_line(line) panel.text(xr-dxr/2-epsxr,yr+epsyr,r'$10$\%',fontsize=8) panel.text(xr-epsxr,yr+epsyr,r'$25$\%',fontsize=8) panel.text(xr+dxr/2-epsxr,yr+epsyr,r'$50$\%',fontsize=8) panel.text(xr+dxr-epsxr,yr+epsyr,r'$75$\%',fontsize=8) panel.text(xr+dxr+dxr/2-epsxr,yr+epsyr,r'$90$\%',fontsize=8) ######################## panel B ######################################## panel = fig.add_axes([x2, y2, dx2, dy2]) panel.tick_params(axis='both',which='both',direction='in',bottom=True,top=True,left=False,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_xlim(xlim_middle,xlim_right) panel.set_ylim(ylim_low,ylim_up) panel.xaxis.set_major_locator(MultipleLocator(200)) panel.xaxis.set_minor_locator(MultipleLocator(50)) panel.yaxis.set_major_locator(MultipleLocator(0.5)) panel.yaxis.set_minor_locator(MultipleLocator(0.1)) panel.set_yticklabels([]) # plot avg line panel.plot(episodes,avg_time,'-',color='darkgreen',markersize=2,linewidth=2,zorder=3) panel.plot(episodes[0:120],benchmark_passive[0:120],'b--',markersize=2,linewidth=1.5,zorder=3) #~ panel.plot(episodes,benchmark_optimal,'k--',markersize=2,linewidth=1.5,zorder=3,label=r'optimal agent (avg.)') #~ panel.plot(episodes,yfunc,'r--',markersize=2,linewidth=0.9,zorder=3) # to uncomment if we wnat to add the fit line # plot boxes c = "green" xwidth = 35 binwidth = 50 for index in range(len(episodes)): if (index%binwidth == 0): x = episodes[index] data = np.zeros((len(times[index]))) for k in range(len(times[index])): data[k]=times[index][k] panel.boxplot(data, sym='', whis=[10,90], positions=[x], widths = xwidth, manage_xticks=False, autorange=False, patch_artist=True,boxprops=dict(facecolor='lightgreen', color=c),capprops=dict(color=c),whiskerprops=dict(color=c),medianprops=dict(color=c)) ######################## 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(axisticslabelfontsizeinset) for tick in panel.yaxis.get_major_ticks(): tick.label.set_fontsize(axisticslabelfontsizeinset) panel.set_xlim(0.5,3300) panel.set_ylim(0.0,1.8) panel.yaxis.set_major_locator(MultipleLocator(0.5)) panel.yaxis.set_minor_locator(MultipleLocator(0.1)) panel.set_xlabel(r'Pe',fontsize=axislabelfontsizeinset) panel.set_xscale('log') panel.set_xticks([1.,10.,100.,1000.]) x_minor = ticker.LogLocator(base=10.0,subs=(0.2,0.4,0.6,0.8),numticks=12) panel.xaxis.set_minor_locator(x_minor) panel.xaxis.set_minor_formatter(ticker.NullFormatter()) panel.text(0.4,0.83,r'$10^3$-th episode',fontsize=axislabelfontsize-3,transform=panel.transAxes) # plot avg line panel.plot(Pes,TRTs,'-',color='saddlebrown',markersize=2,linewidth=1.5,zorder=3,label=r'learning agent') panel.plot(Pes,benchmark_passive_Pes,'b--',markersize=2,linewidth=1.5,zorder=3,label=r'passive particle (avg.)') #~ panel.plot(episodes,benchmark_optimal,'k--',markersize=2,linewidth=1.5,zorder=3,label=r'optimal agent (avg.)') # plot boxes c = "brown" index = len(episodes)-1 factor_bin_width = 2.6 x = 100.0 xwidth = x/factor_bin_width times = times100 data = np.zeros((len(times[index]))) for k in range(len(times[index])): data[k]=times[index][k] panel.boxplot(data, sym='', whis=[10,90], positions=[x], widths = xwidth, manage_xticks=False, autorange=False, patch_artist=True,boxprops=dict(facecolor='peru', color=c),capprops=dict(color=c),whiskerprops=dict(color=c),medianprops=dict(color=c)) x = 20.0 xwidth = x/factor_bin_width times = times20 data = np.zeros((len(times[index]))) for k in range(len(times[index])): data[k]=times[index][k] panel.boxplot(data, sym='', whis=[10,90], positions=[x], widths = xwidth, manage_xticks=False, autorange=False, patch_artist=True,boxprops=dict(facecolor='peru', color=c),capprops=dict(color=c),whiskerprops=dict(color=c),medianprops=dict(color=c)) x = 1.0 xwidth = x/factor_bin_width times = times1 data = np.zeros((len(times[index]))) for k in range(len(times[index])): data[k]=times[index][k] panel.boxplot(data, sym='', whis=[10,90], positions=[x], widths = xwidth, manage_xticks=False, autorange=False, patch_artist=True,boxprops=dict(facecolor='peru', color=c),capprops=dict(color=c),whiskerprops=dict(color=c),medianprops=dict(color=c)) x = 2.0 xwidth = x/factor_bin_width times = times2 data = np.zeros((len(times[index]))) for k in range(len(times[index])): data[k]=times[index][k] panel.boxplot(data, sym='', whis=[10,90], positions=[x], widths = xwidth, manage_xticks=False, autorange=False, patch_artist=True,boxprops=dict(facecolor='peru', color=c),capprops=dict(color=c),whiskerprops=dict(color=c),medianprops=dict(color=c)) x = 5.0 xwidth = x/factor_bin_width times = times5 data = np.zeros((len(times[index]))) for k in range(len(times[index])): data[k]=times[index][k] panel.boxplot(data, sym='', whis=[10,90], positions=[x], widths = xwidth, manage_xticks=False, autorange=False, patch_artist=True,boxprops=dict(facecolor='peru', color=c),capprops=dict(color=c),whiskerprops=dict(color=c),medianprops=dict(color=c)) x = 10.0 xwidth = x/factor_bin_width times = times10 data = np.zeros((len(times[index]))) for k in range(len(times[index])): data[k]=times[index][k] panel.boxplot(data, sym='', whis=[10,90], positions=[x], widths = xwidth, manage_xticks=False, autorange=False, patch_artist=True,boxprops=dict(facecolor='peru', color=c),capprops=dict(color=c),whiskerprops=dict(color=c),medianprops=dict(color=c)) x = 1000.0 xwidth = x/factor_bin_width times = times1000 data = np.zeros((len(times[index]))) for k in range(len(times[index])): data[k]=times[index][k] panel.boxplot(data, sym='', whis=[10,90], positions=[x], widths = xwidth, manage_xticks=False, autorange=False, patch_artist=True,boxprops=dict(facecolor='peru', color=c),capprops=dict(color=c),whiskerprops=dict(color=c),medianprops=dict(color=c)) x = 500.0 xwidth = x/factor_bin_width times = times500 data = np.zeros((len(times[index]))) for k in range(len(times[index])): data[k]=times[index][k] panel.boxplot(data, sym='', whis=[10,90], positions=[x], widths = xwidth, manage_xticks=False, autorange=False, patch_artist=True,boxprops=dict(facecolor='peru', color=c),capprops=dict(color=c),whiskerprops=dict(color=c),medianprops=dict(color=c)) x = 50.0 xwidth = x/factor_bin_width times = times50 data = np.zeros((len(times[index]))) for k in range(len(times[index])): data[k]=times[index][k] panel.boxplot(data, sym='', whis=[10,90], positions=[x], widths = xwidth, manage_xticks=False, autorange=False, patch_artist=True,boxprops=dict(facecolor='peru', color=c),capprops=dict(color=c),whiskerprops=dict(color=c),medianprops=dict(color=c)) x = 200.0 xwidth = x/factor_bin_width times = times200 data = np.zeros((len(times[index]))) for k in range(len(times[index])): data[k]=times[index][k] panel.boxplot(data, sym='', whis=[10,90], positions=[x], widths = xwidth, manage_xticks=False, autorange=False, patch_artist=True,boxprops=dict(facecolor='peru', color=c),capprops=dict(color=c),whiskerprops=dict(color=c),medianprops=dict(color=c)) x = 2000.0 xwidth = x/factor_bin_width times = times2000 data = np.zeros((len(times[index]))) for k in range(len(times[index])): data[k]=times[index][k] panel.boxplot(data, sym='', whis=[10,90], positions=[x], widths = xwidth, manage_xticks=False, autorange=False, patch_artist=True,boxprops=dict(facecolor='peru', color=c),capprops=dict(color=c),whiskerprops=dict(color=c),medianprops=dict(color=c)) pdf.savefig(fig) return def main(): N_bunches = 1 foldername = '../RESULTS/Pe100/' N_runs = 5000 # number of completely independent runs N_episodes = 1000 # number of episodes. Each episode lasts for a time_single_episode avg_time_passive = 1.14166 avg_time_optimal = 0.28629 # for Pe=100 from "Adaptive active Brownian particles searching for targets of unknown positions" #~ avg_time_optimal = 0.74484 # for Pe=20 from "Adaptive active Brownian particles searching for targets of unknown positions" # read average target times filename = foldername+'AvgTargetTimes_bunch' episodes,avg_time = read_AvgTargetTimes_bunches(filename,N_bunches,N_episodes) # read target times filename = foldername+'TargetTimes_bunch' times = read_TargetTimes_bunches(filename,N_bunches,N_runs,N_episodes) # fit test A0 = avg_time[0] Ainf = avg_time[-1] yfit = [] for t in range(len(avg_time)): yfit.append((avg_time[t]-Ainf)/(A0-Ainf)) popt, pcov = curve_fit(func, episodes, yfit, p0=[1.0]) perr = np.sqrt(np.diag(pcov)) print (popt) print (perr) yfunc = [] for t in range(len(avg_time)): yfunc.append( func(episodes[t],popt[0])*(A0-Ainf) + Ainf ) # for inset Pes=np.asarray([1,2,5,10,20,50,100,200,500,1000,2000]) times100 = times avgtime100 = avg_time[-1] foldername = '../RESULTS/Pe20/' filename = foldername + 'TargetTimes_bunch' times20 = read_TargetTimes_bunches(filename,1,N_runs,N_episodes) filename = foldername+'AvgTargetTimes_bunch' episodes20,avg_time20 = read_AvgTargetTimes_bunches(filename,1,N_episodes) avgtime20 = avg_time20[-1] foldername = '../RESULTS/Pe1_Kaur/' filename = foldername + 'TargetTimes_bunch' times1 = read_TargetTimes_bunches(filename,1,N_runs,N_episodes) filename = foldername+'AvgTargetTimes_bunch' episodes1,avg_time1 = read_AvgTargetTimes_bunches(filename,1,N_episodes) avgtime1 = avg_time1[-1] foldername = '../RESULTS/Pe2/' filename = foldername + 'TargetTimes_bunch' times2 = read_TargetTimes_bunches(filename,2,N_runs/2,N_episodes) filename = foldername+'AvgTargetTimes_bunch' episodes2,avg_time2 = read_AvgTargetTimes_bunches(filename,2,N_episodes) avgtime2 = avg_time2[-1] foldername = '../RESULTS/Pe5/' filename = foldername + 'TargetTimes_bunch' times5 = read_TargetTimes_bunches(filename,2,N_runs/2,N_episodes) filename = foldername+'AvgTargetTimes_bunch' episodes5,avg_time5 = read_AvgTargetTimes_bunches(filename,2,N_episodes) avgtime5 = avg_time5[-1] foldername = '../RESULTS/Pe10/' filename = foldername + 'TargetTimes_bunch' times10 = read_TargetTimes_bunches(filename,2,N_runs/2,N_episodes) filename = foldername+'AvgTargetTimes_bunch' episodes10,avg_time10 = read_AvgTargetTimes_bunches(filename,2,N_episodes) avgtime10 = avg_time10[-1] foldername = '../RESULTS/Pe1000/' filename = foldername + 'TargetTimes_bunch' times1000 = read_TargetTimes_bunches(filename,1,N_runs,N_episodes) filename = foldername+'AvgTargetTimes_bunch' episodes1000,avg_time1000 = read_AvgTargetTimes_bunches(filename,1,N_episodes) avgtime1000 = avg_time1000[-1] foldername = '../RESULTS/Pe50/' filename = foldername + 'TargetTimes_bunch' times50 = read_TargetTimes_bunches(filename,1,N_runs,N_episodes) filename = foldername+'AvgTargetTimes_bunch' episodes50,avg_time50 = read_AvgTargetTimes_bunches(filename,1,N_episodes) avgtime50 = avg_time50[-1] foldername = '../RESULTS/Pe500/' filename = foldername + 'TargetTimes_bunch' times500 = read_TargetTimes_bunches(filename,1,N_runs,N_episodes) filename = foldername+'AvgTargetTimes_bunch' episodes500,avg_time500 = read_AvgTargetTimes_bunches(filename,1,N_episodes) avgtime500 = avg_time500[-1] foldername = '../RESULTS/Pe200/' filename = foldername + 'TargetTimes_bunch' times200 = read_TargetTimes_bunches(filename,1,N_runs,N_episodes) filename = foldername+'AvgTargetTimes_bunch' episodes200,avg_time200 = read_AvgTargetTimes_bunches(filename,1,N_episodes) avgtime200 = avg_time200[-1] foldername = '../RESULTS/Pe2000/' filename = foldername + 'TargetTimes_bunch' times2000 = read_TargetTimes_bunches(filename,5,N_runs/5,N_episodes) filename = foldername+'AvgTargetTimes_bunch' episodes2000,avg_time2000 = read_AvgTargetTimes_bunches(filename,1,N_episodes) avgtime2000 = avg_time2000[-1] TRTs=np.asarray([avgtime1,avgtime2,avgtime5,avgtime10,avgtime20,avgtime50,avgtime100,avgtime200,avgtime500,avgtime1000,avgtime2000]) # plots make_figure(episodes,avg_time,avg_time_passive,avg_time_optimal,times,yfunc,Pes,TRTs,times1,times2,times5,times10,times20,times50,times100,times200,times500,times1000,times2000) main()