from pylab import * import csv from jmoo_problems import * from jmoo_algorithms import * from jmoo_properties import * from utility import * fignum = 0 data = [] heads = [] best = [] foam = [] RRS_scores = [] RRS = [] for p,prob in enumerate(problems): data.append([]) heads.append([]) best.append([]) foam.append([]) RRS.append([]) RRS_scores.append([]) for a,alg in enumerate(algorithms): finput = open("data/results_" + prob.name + "_" + alg.name + ".datatable", 'rb') f2input = open(DATA_PREFIX + RRS_TABLE + "_" + prob.name + "_" + alg.name + DATA_SUFFIX, 'rb') reader = csv.reader(finput, delimiter=',') reader2 = csv.reader(f2input, delimiter=',') data[p].append( [] ) heads[p].append( [] ) best[p].append( [] ) foam[p].append( [] ) RRS[p].append( [] ) RRS_scores[p].append( [] ) for o,obj in enumerate(prob.objectives): foam[p][a].append([]) foam[p][a][o] = {} RRS[p][a].append([]) RRS[p][a][o] = {} for i,row in enumerate(reader2): for j,col in enumerate(row): if i == 0: print col RRS_scores[p][a].append([float(col.strip("%)( "))]) else: RRS_scores[p][a][j].append(float(col.strip("%)( "))) for o,obj in enumerate(prob.objectives): n = RRS_scores[p][a][-1][-1] n = (int(round(n/5.0)*5.0)) if n in RRS[p][a][o]: RRS[p][a][o][n].append(float(RRS_scores[p][a][o][-1])) else: RRS[p][a][o][n] = [float(RRS_scores[p][a][o][-1])] for i,row in enumerate(reader): if not str(row[0]) == "0": for j,col in enumerate(row): if i == 0: heads[p][a].append(col) data[p][a].append([]) best[p][a].append(999999) else: if not col == "": data[p][a][j].append(float(col.strip("%)("))) if data[p][a][j][-1] < best[p][a][j]: best[p][a][j] = data[p][a][j][-1] # row is now read if i > 0: for o,obj in enumerate(prob.objectives): n = data[p][a][0][-1] n = (int(round(n/20.0)*20.0)) if n in foam[p][a][o]: foam[p][a][o][n].append(float(data[p][a][o*3+2][-1])) else: foam[p][a][o][n] = [float(data[p][a][o*3+2][-1])] # all rows read # Interpolate foam keys for doit in range(5): for o,obj in enumerate(prob.objectives): keys = sorted(foam[p][a][o].keys()) for key,nextkey in zip(keys[0:-1], keys[1:]): foam[p][a][o][(nextkey-key)/2 + key] = foam[p][a][o][key] keys = sorted(foam[p][a][o].keys()) """ if alg.name == "GALE": skips = 5 dur = 20000 else: skips = 100 dur = 20000 for o,obj in enumerate(prob.objectives): for n in range(0, dur): nk = (int(round(n/float(skips))*float(skips))) if nk >= dur: nk = dur - skips nk1 = nk nk2 = nk while len(foam[p][a][o][n]) == 0: # and (nk1 < dur or nk2 >= 0): if nk1 < dur and len(foam[p][a][o][n]) == 0 and len(foam[p][a][o][nk1]) > 0: for na in range(n, nk1): foam[p][a][o][na].append(min(foam[p][a][o][nk1]) if obj.lismore else max(foam[p][a][o][nk1])) if nk2 >= 0 and len(foam[p][a][o][n]) == 0 and len(foam[p][a][o][nk2]) > 0: for na in range(nk2+1, n+1): foam[p][a][o][na].append(min(foam[p][a][o][nk2]) if obj.lismore else max(foam[p][a][o][nk2])) nk1 = nk1 + skips #nk1 = (int(round(nk1/float(5.0))*float(5.0))) nk2 = nk2 - skips #nk2 = (int(round(nk2/float(5.0))*float(5.0))) for n in range(0, 20000): # print n, len(foam[p][a][o][n]) if len(foam[p][a][o][n]) == 0: foam[p][a][o][n].append(prob.referencePoint[o]) """ fignum = 0 colors = ['r', 'b', 'g'] from matplotlib.font_manager import FontProperties font = {'family' : 'sans-serif', 'weight' : 'normal', 'size' : 8} matplotlib.rc('font', **font) fontP = FontProperties() fontP.set_size('x-small') codes = ["rx", "b+", "g1"] codes2= ["r-", "b-", "g-"] colors= ["r", "b", "g"] symbols=["x", "+", "1"] line = "-" dotted= "--" algnames = [alg.name for alg in algorithms] axy = [0,1,2,3] axx = [0,0,0,0] f, axarr = plt.subplots(max(2,max([len(prob.objectives) for prob in problems])), max(2,len(problems))) for p,prob in enumerate(problems): for o,obj in enumerate(prob.objectives): maximum = -9999999.9 minimum = +9999999.9 if not o == 112: o_o = o if o > 4: o_o = o - 1 for a,alg in enumerate(algorithms): maximum = max(max(data[p][a][o*3+2]), maximum) minimum = min(min(data[p][a][o*3+2]), minimum) for a,alg in enumerate(algorithms): if o == 0: axarr[o_o][p].set_title(prob.name) if p == 0: axarr[o_o][p].set_ylabel(prob.objectives[o].name, rotation=90) #axarr[o_o][p].plot([x for x in range(0,10000,10)], [100 for x in range(0,10000,10)], 'k-', [int(round(x/5.0)*5.0) for x in data[p][a][0]], data[p][a][o*3+2], codes[a],label=alg.name, markersize=1) #axarr[o_o][p].plot([x for x in range(0,10000,10)], [best[p][a][o*3+2] for x in range(0,10000,10)], colors[a]+dotted, markersize=1) #axarr[o_o][p].plot([x for x in data[p][a][0]], [best[p][a][o*3+2] for x in data[p][a][0]], codes2[a]) #axarr[o_o][p].plot([x for x in range(0,10000,10)], [100 for x in range(0,10000,10)], 'k-', [key for key in RRS[p][a][o].keys()], [getPercentile(RRS[p][a][o], 25), getPercentile(RRS[p][a][o], 50), getPercentile(RRS[p][a][o], 75)], codes[a],label=alg.name, markersize=4) X = [] Y = [] for key in RRS[p][a][o]: #print key, getPercentile(RRS[p][a][o][key], 25), getPercentile(RRS[p][a][o][key], 50), getPercentile(RRS[p][a][o][key], 75) X += [key, key, key] Y += [getPercentile(RRS[p][a][o][key], 25), getPercentile(RRS[p][a][o][key], 50), getPercentile(RRS[p][a][o][key], 75)] for r in range(len(X)/3): axarr[o_o][p].plot(X[r*3:(r+1)*3], Y[r*3:(r+1)*3], codes2[a],label=alg.name, markersize=1) for r in range(3): axarr[o_o][p].plot([X[(k)*3+r] for k in range(len(X)/3)], [Y[(k)*3+r] for k in range(len(X)/3)], colors[a]+["_", "o", "_"][r % 3],label=alg.name, markersize=[8, 3, 8][r]) axarr[o_o][p].plot([x for x in range(0,10000,10)], [100 for x in range(0,10000,10)], 'k-') #axarr[o_o][p].plot([key for key in sorted(foam[p][a][o].keys())], [min(foam[p][a][o][key]) for key in sorted(foam[p][a][o].keys())], colors[a]+"-", markersize=1) axarr[o_o][p].set_autoscale_on(False) axarr[o_o][p].set_xlim([0, 10000]) axarr[o_o][p].set_xscale('log') if maximum > 1000: axarr[o_o][p].set_yscale('log') f.set_size_inches(2.75*4, 2.75*3) """ if obj.name == "Cost": axarr[o_o][p].set_ylim([-10, 210]) elif obj.name == "Completion": axarr[o_o][p].set_ylim([50, 150]) elif obj.name == "Idle": axarr[o_o][p].set_ylim([-10, 210]) elif obj.name == "Score": axarr[o_o][p].set_ylim([-10, 140]) else: """ axarr[o_o][p].set_ylim([minimum-(maximum-minimum)*0.10, maximum+(maximum-minimum)*0.10]) #if prob.name == "Viennet2": # axarr[o_o][p].set_ylim([50, 110]) #xlabel("num evaluations") #ylabel("% of initial") #suptitle(prob.name, fontsize=11) plt.subplots_adjust(top=0.85) plt.subplots_adjust(bottom=0.15) legend(loc='upper center', bbox_to_anchor=(-0.10, -0.10), ncol=3, prop=fontP) #show() fignum = len([name for name in os.listdir('charts/' + date_folder_prefix)]) print fignum plt.savefig('charts/' + date_folder_prefix + '/figure' + str("%02d" % fignum) + "_" + "objectives_" + prob.name + "_" + tag + '.png', dpi=100) cla()