def plot_what_i_want(pfcmd, weights, rates, config):
PFCrates, MDinputs, MDrates, Outrates, Inputs, Targets, MSEs= rates
wOuts, wPFC2MDs, wMD2PFCs, wMD2PFCMults, wJrecs, MDpreTraces = weights
# these tensors are training_i x tsteps x no_neuron
p = config.Nsub//2
tpb = config.trials_per_block
Ntrain = PFCrates[:,:, :5].shape[0]
yticks = (0, 0.5,1)
xticks = [0, 1000, 2000]
pfcmd.figCustom, axes = plt.subplots(4,3)#, sharex=True)# , sharey=True)
pfcmd.figCustom.set_size_inches([9,7])
plot_trials = [0, 1, 2, 9, tpb, tpb+8, tpb+9, tpb+10, tpb*2, tpb*2+8, tpb*2+10]
faxes = axes.flatten()
for xi, ai in enumerate(plot_trials):
ax = faxes[xi]
ax.hist(MDpreTraces[ai,p*0:p*0+p].flatten(), alpha=0.7 )#, 'tab:blue') # take a slice from context 1 #[traini, tstep, Nneur]
ax.hist(MDpreTraces[ai,p*1:p*1+p].flatten(), alpha=0.7 )#, 'tab:blue') # take a slice from context 1 #[traini, tstep, Nneur]
pltu.axes_labels(ax, 'MDpre trial {}'.format(ai), 'freq')
plt.text(0.01, 0.1, str(Inputs[ai])+ str(Targets[ai]), transform=ax.transAxes)
ax.set_ylim([0, 15])
ax.set_xlim([0, .7])
def plot_rates(pfcmd, rates, config):
PFCrates, MDinputs, MDrates, Outrates, Inputs, Targets, MSEs= rates
# these tensors are training_i x tsteps x no_neuron
p = config.Nsub//2
tpb = config.trials_per_block
Ntrain = PFCrates[:,:, :5].shape[0]
yticks = (0, 0.5,1)
xticks = [0, 1000, 2000]
pfcmd.figRates, axes = plt.subplots(4,3)#, sharex=True)# , sharey=True)
pfcmd.figRates.set_size_inches([9,7])
ax = axes[0,0]
ax.plot(range(Ntrain),np.mean( PFCrates[:,:,:5], axis=1), '.', markersize =0.5)
# ax.plot(range(Ntrain), np.mean( PFCrates[:, :,:p] , axis=(1,2)), '-', linewidth=-.5)
pltu.beautify_plot(ax,x0min=False,y0min=False, yticks=yticks, xticks=xticks)
pltu.axes_labels(ax,'','Mean FR')
ax.set_ylim([0,1])
ax.set_title('PFC Up-V1')
ax = axes[0,1]
ax.plot(range(Ntrain),np.mean( PFCrates[:, :,p:p+5], axis=1), '.', markersize =0.5)
# ax.plot(range(Ntrain), np.mean( PFCrates[:, :,p:p*2] , axis=(1,2)), '-', linewidth=-.5)
pltu.beautify_plot(ax,x0min=False,y0min=False, yticks=yticks, xticks=xticks)
pltu.axes_labels(ax,'','')
ax.set_ylim([0,1])
ax.set_title('PFC Up-V2')
ax = axes[0,2]
ax.plot(range(Ntrain),np.mean( PFCrates[:, :,p*2:p*2+5], axis=1), '.', markersize =0.5)
# ax.plot(range(Ntrain), np.mean( PFCrates[:, :,p*2:p*3] , axis=(1,2)), '-', linewidth=0.5)
pltu.beautify_plot(ax,x0min=False,y0min=False)
ax.set_ylim([0,1])
pltu.axes_labels(ax,'','')
ax.set_title('PFC Down-V1')
ninp = np.array(Inputs)
ax = axes[1,0]
#plot trials with up cue or down cue with blue or red.
ax.plot(np.arange(0,Ntrain)[ninp[:,0]==1.],np.mean( MDrates[:,:,0][ninp[:,0]==1.], axis=1), '.', markersize =0.5, color='tab:blue', label='Up')
ax.plot(np.arange(0,Ntrain)[ninp[:,0]==0.],np.mean( MDrates[:,:,0][ninp[:,0]==0.], axis=1), '.', markersize =0.5, color='tab:red', label='Down')
ax.legend()
# ax.plot(range(Ntrain),np.mean( MDrates[:,:,0], axis=1), '.', markersize =0.5)
pltu.beautify_plot(ax,x0min=False,y0min=False, yticks=yticks, xticks=xticks)
pltu.axes_labels(ax,'','mean FR')
ax.set_title('MD 0')
if (len(config.variable_trials_per_block) > 0):
for ib in range(len(config.variable_trials_per_block)-1):
xmin = config.variable_trials_per_block[ib]
xmax = config.variable_trials_per_block[ib+1]
ax.axvspan(xmin, xmax, alpha=0.1, color='grey')
else:
for ib in range(1, config.Nblocks,2):
ax.axvspan(tpb* ib, tpb*(ib+1), alpha=0.1, color='grey')
ax = axes[1,1]
ax.plot(range(Ntrain),np.mean( MDrates[:,:,1], axis=1), '.', markersize =0.5)
pltu.beautify_plot(ax,x0min=False,y0min=False, yticks=yticks, xticks=xticks)
pltu.axes_labels(ax,'','')
ax.set_title('MD 1')
ax = axes[1,2]
ax.plot(range(Ntrain),np.mean( MDinputs[:, :,:], axis=1), '.', markersize =0.5)
pltu.beautify_plot(ax,x0min=False,y0min=False, xticks=xticks)
pltu.axes_labels(ax,'','')
ax.set_title('MD avg inputs')
if (len(config.variable_trials_per_block) > 0):
for ib in range(len(config.variable_trials_per_block)-1):
xmin = config.variable_trials_per_block[ib]
xmax = config.variable_trials_per_block[ib+1]
ax.axvspan(xmin, xmax, alpha=0.1, color='grey')
else:
for ib in range(1, config.Nblocks,2):
ax.axvspan(tpb* ib, tpb*(ib+1), alpha=0.1, color='grey')
# ax = axes[2,0]
# ax.plot(range(Ntrain),np.mean( Outrates[:,:,0], axis=1), '.', markersize =0.5)
# pltu.beautify_plot(ax,x0min=False,y0min=False, yticks=yticks, xticks=xticks)
# pltu.axes_labels(ax,'','mean FR')
# ax.set_title('Out 0')
# ax = axes[2,1]
# ax.plot(range(Ntrain),np.mean( Outrates[:,:,1], axis=1), '.', markersize =0.5)
# pltu.beautify_plot(ax,x0min=False,y0min=False, yticks=yticks, xticks=xticks)
# pltu.axes_labels(ax,'','')
# ax.set_title('Out 1')
# ax = axes[2,2]
# ax.plot(range(Ntrain),np.mean( Outrates[:, :,:], axis=1), '.', markersize =0.5)
# pltu.beautify_plot(ax,x0min=False,y0min=False, yticks=yticks, xticks=xticks)
# pltu.axes_labels(ax,'','')
# ax.set_title('Out 0 and 1')
ax = axes[3,0]
# Plot MSE
ax.plot(MSEs)
ax.plot(smooth(MSEs, 8), 'tab:orange', linewidth= pltu.linewidth)
pltu.beautify_plot(ax,x0min=False,y0min=False, xticks=xticks)
pltu.axes_labels(ax,'Trials','MSE')
ax.set_title('MSE')
# ax.plot(range(Ntrain),Inputs[:, 0] + np.random.uniform(-0.1, 0.1, size=(Ntrain,1)) , 'o', markersize =0.5)
# ax.plot(range(Ntrain),Targets[:, 0] + np.random.uniform(-0.1, 0.1, size=(Ntrain,1)) , 'o', markersize =0.5)
# pltu.beautify_plot(ax,x0min=False,y0min=False)
# pltu.axes_labels(ax,'','Inputs')
# pfcmd.figOuts = plt.figure()
# ax = pfcmd.figOuts.add_subplot(311)
# ax.plot(range(Ntrain), 1.*(Targets[:,0] == out_higher_mean) -0.3+ np.random.uniform(-0.01, 0.01, size=(Ntrain,) ) , '.', markersize =0.5)
# ax.set_title('Percent correct answers smoothened over 20 trials')
# ax = pfcmd.figOuts.add_subplot(312)
# ax.plot(smooth((Targets[:,0] == out_higher_mean)*1., 20), linewidth=pltu.linewidth)
# pltu.axes_labels(ax, 'Trials', '% Correct')
# out_higher_endFR =1.*( Outrates[:, -1 ,0] > Outrates[:, -1 ,1] )
out_higher_mean = 1.*( np.mean( Outrates[:, :,0], axis=1) > np.mean( Outrates[:, :,1], axis=1) )
Matches = 1. * (Targets[:,0] == Inputs[:,0]) # Targets is [n_trials x 2] Inputs is N_trials x 4 (cue and q values)
Responses= 1.* (out_higher_mean == Inputs[:,0]) #* 0.8 + 0.1 #+ np.random.uniform(-noise, noise, size=(Ntrain,) )
Corrects = 1. * (Targets[:,0] == out_higher_mean)
Matches = Matches *1.6-0.28
Responses = Responses *1.2-0.1
stages = 4
no_trials_to_score = 100
if len(config.variable_trials_per_block) > 0:
tpb = 10 # NOTE: This is a hack just to get the code to run with variable
pfcmd.score = [np.mean(Corrects[istage*tpb:(istage*tpb)+no_trials_to_score])* 100. for istage in range(1, stages+1)] # score binnged into stages
pfcmd.score.append(np.mean(pfcmd.score[:-1])) # The avrg of the cognitive flex measures, except the last forced switch block.
pfcmd.score.append(np.mean(Corrects) * 100. ) # Add a var that holds the score of the model. % correct response. Later to be outputed as a text file.
pfcmd.corrects = Corrects
noise = 0.15
ax = axes[3,1]
ax.plot(Matches + np.random.uniform(-noise, noise, size=(Ntrain,) ), 'o', markersize = 0.25, alpha=0.7)
ax.plot(Responses+ np.random.uniform(-noise, noise, size=(Ntrain,) ), 'o', markersize = 0.25, alpha=0.7)
pltu.axes_labels(ax, 'Trials', 'non-match Match')
# ax.set_title('Blue o: Correct Orange x: response')
ax.set_ylim([-0.3, 1.3])
# ax.set_xlim([0, 2200])
ax = axes[3,2] # Firing rates distribution
# print('Shape is: ', PFCrates.shape)
ax.hist(PFCrates[900:1000].flatten(), alpha=0.7 ) #, 'tab:blue') # take a slice from context 1 #[traini, tstep, Nneur]
ax.hist(PFCrates[2000:2100].flatten(), alpha= 0.5) #, 'tab:red') # context 0
pltu.axes_labels(ax, 'rates', 'freq')
# PLOT BEHAVIOR MEASURES
pfcmd.figOuts = plt.figure()
pfcmd.figOuts.set_size_inches([9,7])
noise = 0.07
ax = pfcmd.figOuts.add_subplot(211)
ax.plot(Matches + np.random.uniform(-noise, noise, size=(Ntrain,) ), 'o', markersize = 0.25, alpha=0.8)
ax.plot(Responses+ np.random.uniform(-noise, noise, size=(Ntrain,) ), 'o', markersize = 0.25, alpha=0.8)
pltu.axes_labels(ax, 'Trials', 'non-match V1 Match')
ax.set_title('Blue: Correct Orange: response')
ax.set_ylim([-0.8, 1.8])
rm = np.convolve(Corrects, np.ones((40,))/40, mode='same')
ax.plot(rm, color='black', linewidth= 0.5, alpha = 0.8)
ax.plot(Inputs[:,2], color='tab:red', alpha=0.7, linewidth=0.5)
for bi in range(config.Nblocks):
plt.text((1/(config.Nblocks+1))* (0.74+bi), 0.1, str(config.block_schedule[bi]), transform=ax.transAxes)
ax = pfcmd.figOuts.add_subplot(212)
ax.plot(Inputs[:,4]*.1, color='tab:red', alpha=0.7, linewidth=0.5, label='cx=match')
st = tpb*min(4, config.Nblocks-1) - 10
d = 30
ax.plot(range(st, st+d), Inputs[st:st+d,5], 'o', markersize= 2, linewidth=0.5, color='tab:blue', alpha=0.7, label='sw_dots')
ax.plot(Inputs[:,5], color='tab:green', alpha=0.7, linewidth=0.5, label='p(sw)')
ax.plot(Inputs[:,6], color='tab:blue', alpha=0.5, linewidth=0.5, label='p(r)')
for bi in range(config.Nblocks): # LABEL contexts
plt.text((1/13)* (0.74+bi), 0.1, str(config.block_schedule[bi]), transform=ax.transAxes)
for ib in range(1, config.Nblocks,2): # demarcate contexts with grey shading
ax.axvspan(tpb* ib, tpb*(ib+1), alpha=0.1, color='grey')
ax.legend()
# ax.plot(Matches, 'o', markersize = 3)
# ax.plot(Responses, 'x', markersize = 3)
# pltu.axes_labels(ax, 'Trials', 'non-match Match')
# ax.set_ylim([-0.3, 1.3])
# ax.set_xlim([1970, 2050])
plt.text(0.01, -0.1, str(config.args_dict), transform=ax.transAxes)
# fig, axx = plt.subplots(3,1)
# ax = axx[0]
# t = tpb*min(3, config.Nblocks-1) - 10
# d = 30
# ax.plot(range(t, t+d), Inputs[t:t+d,6], 'o', markersize= 1, linewidth=0.5, color='tab:blue', alpha=0.7, label='sm_dots')
# ax = axx[1]
# ax.plot(range(t, t+d), Inputs[t:t+d,6], 'o', markersize= 1, linewidth=0.5, color='tab:blue', alpha=0.7, label='sm_dots')
# ax = axx[2]
# d = 50
# ax.plot(range(t, t+d), Inputs[t:t+d,6], 'o', markersize= 1, linewidth=0.5, color='tab:blue', alpha=0.7, label='sm_dots')
# fig.savefig('./results/switch_signal.png')
return # NOTE: Sabrina trying to hack
pfcmd.figRates
pfcmd.figRates.tight_layout()
# PLOT within trial activity for 4 selected trials:
trials_to_draw = [0,config.trials_per_block, config.trials_per_block+100]# [0, config.trials_per_block, int(config.Nblocks//4*config.trials_per_block)]
pfcmd.figTrials, axes = plt.subplots(5,len(trials_to_draw))#, sharex=True)# , sharey=True)
pfcmd.figTrials.set_size_inches([9,3*len(trials_to_draw)])
for ti, trial in enumerate(range(len(trials_to_draw))):
ax = axes[0,ti]
ax.plot(range(200),np.mean( PFCrates[trial,:,:p], axis=1), '-', linewidth=1)
ax.plot(range(200), PFCrates[trial,:,:5], '-', linewidth=0.5)
pltu.beautify_plot(ax,x0min=False,y0min=False, yticks=yticks, xticks=xticks)
pltu.axes_labels(ax,str(Inputs[trial]),'PFC Up-V1')
ax.set_title('PFC Up-V1')
ax = axes[1,ti]
ax.plot(range(200),np.mean( PFCrates[trial,:,p:p*2], axis=1), '-', linewidth=1)
ax.plot(range(200), PFCrates[trial,:,p:p+5], '-', linewidth=0.5)
pltu.beautify_plot(ax,x0min=False,y0min=False, yticks=yticks, xticks=xticks)
pltu.axes_labels(ax,str(Targets[trial]),'Up-V2')
# ax.set_title('PFC Up-V2')
ax = axes[2,ti]
ax.plot(range(200),np.mean( PFCrates[trial,:,p*2:p*3], axis=1), '-', linewidth=1)
ax.plot(range(200), PFCrates[trial,:,2*p:2*p+5], '-', linewidth=0.5)
pltu.beautify_plot(ax,x0min=False,y0min=False, yticks=yticks, xticks=xticks)
pltu.axes_labels(ax,'','Down-V1')
# ax.set_title('PFC Down-V1')
ax = axes[3,ti]
ax.plot(range(200), MDrates[trial,:,:], '-', linewidth=1, alpha=0.7)
ax.plot(range(200), MDinputs[trial,:,:], '-.', linewidth=2, alpha=0.7)
pltu.beautify_plot(ax,x0min=False,y0min=False, yticks=yticks, xticks=xticks)
pltu.axes_labels(ax,'','FR')
ax.set_title('MD 0 and 1')
ax = axes[4,ti]
ax.plot(range(200), Outrates[trial,:,:], '-', linewidth=1, alpha=0.7)
pltu.beautify_plot(ax,x0min=False,y0min=False, yticks=yticks, xticks=xticks)
pltu.axes_labels(ax,'ms','FR')
ax.set_title('Out 0 and 1')
def plot_weights(pfcmd, weights, config):
wOuts, wPFC2MDs, wMD2PFCs, wMD2PFCMults, wJrecs, MDpreTraces = weights
xticks = [0, 1000, 2000, 3000, 4000]
# plot output weights evolution
pfcmd.figWeights, axes = plt.subplots(5,3)#, sharex=True) #, sharey=True)
# pfcmd.figWeights.set_figheight = pltu.twocolumnwidth
# pfcmd.figWeights.set_figwidth = pltu.twocolumnwidth
pfcmd.figWeights.set_size_inches([9,9])
plot_cue_v_subpop = True
tpb = config.trials_per_block
if plot_cue_v_subpop:
subplot_titles = ['Up-V1', 'Up-V2', 'Down-V1']
p = config.Nsub//2
else:
subplot_titles = ['PFC cue 1', 'PFC cue 2', 'PFC cue 3']
p = config.Nsub
for pi, PFC in enumerate(subplot_titles):
ax = axes[0,pi]
ax.plot(wOuts[:,0, p*pi:p*pi+5],'tab:red', linewidth= pltu.linewidth)
ax.plot(wOuts[:,1, p*pi:p*pi+5],'tab:blue', linewidth= pltu.linewidth)
wmean = np.mean(wOuts[:,1,p*pi:p*pi+p], axis=1)
wstd = np.mean(wOuts[:,1,p*pi:p*pi+p], axis=1)
ax.plot(range(len(wmean)), wmean)
ax.fill_between(range(len(wmean)), wmean-wstd, wmean+wstd, alpha=.4)
pltu.beautify_plot(ax,x0min=False,y0min=False, xticks=xticks)
if pi == 0: pltu.axes_labels(ax,'','to Out-0 & 1 (r,b)')
ax.set_title(PFC)
if (len(config.variable_trials_per_block) > 0):
for ib in range(len(config.variable_trials_per_block)-1):
xmin = config.variable_trials_per_block[ib]
xmax = config.variable_trials_per_block[ib+1]
ax.axvspan(xmin, xmax, alpha=0.1, color='grey')
else:
for ib in range(1, config.Nblocks,2):
ax.axvspan(tpb* ib, tpb*(ib+1), alpha=0.1, color='grey')
for pi, PFC in enumerate(subplot_titles):
ax = axes[1,pi]
ax.plot(wPFC2MDs[:,0, p*pi:p*pi+5],'tab:red', linewidth= pltu.linewidth)
ax.plot(wPFC2MDs[:,1, p*pi:p*pi+5],'tab:blue', linewidth= pltu.linewidth)
wmean = np.mean(wPFC2MDs[:,1,p*pi:p*pi+p], axis=1)
wstd = np.mean(wPFC2MDs[:,1,p*pi:p*pi+p], axis=1)
ax.plot(range(len(wmean)), wmean)
ax.fill_between(range(len(wmean)), wmean-wstd, wmean+wstd, alpha=.4)
pltu.beautify_plot(ax,x0min=False,y0min=False, xticks=xticks)
if pi == 0: pltu.axes_labels(ax,'','to MD-0(r) 1(b)')
if (len(config.variable_trials_per_block) > 0):
for ib in range(len(config.variable_trials_per_block)-1):
xmin = config.variable_trials_per_block[ib]
xmax = config.variable_trials_per_block[ib+1]
ax.axvspan(xmin, xmax, alpha=0.1, color='grey')
else:
for ib in range(1, config.Nblocks,2):
ax.axvspan(tpb* ib, tpb*(ib+1), alpha=0.1, color='grey')
ax = axes[2,pi]
ax.plot(wMD2PFCs[:,p*pi:p*pi+5, 0],'tab:red', linewidth= pltu.linewidth)
ax.plot(wMD2PFCs[:,p*pi:p*pi+5, 1],'tab:blue', linewidth= pltu.linewidth)
pltu.beautify_plot(ax,x0min=False,y0min=False, xticks=xticks)
if pi == 0: pltu.axes_labels(ax,'','from MD-0(r) 1(b)')
if (len(config.variable_trials_per_block) > 0):
for ib in range(len(config.variable_trials_per_block)-1):
xmin = config.variable_trials_per_block[ib]
xmax = config.variable_trials_per_block[ib+1]
ax.axvspan(xmin, xmax, alpha=0.1, color='grey')
else:
for ib in range(1, config.Nblocks,2):
ax.axvspan(tpb* ib, tpb*(ib+1), alpha=0.1, color='grey')
# plot PFC to MD pre Traces
ax = axes[3,pi]
ax.plot(MDpreTraces[:,p*pi:p*pi+5], linewidth = pltu.linewidth)
ax.plot(config.MDlearningBiasFactor*np.mean(MDpreTraces, axis=1), '-.', linewidth = 2)
pltu.beautify_plot(ax,x0min=False,y0min=False, xticks=xticks)
pltu.axes_labels(ax,'Trials','pre')
if (len(config.variable_trials_per_block) > 0):
for ib in range(len(config.variable_trials_per_block)-1):
xmin = config.variable_trials_per_block[ib]
xmax = config.variable_trials_per_block[ib+1]
ax.axvspan(xmin, xmax, alpha=0.1, color='grey')
else:
for ib in range(1, config.Nblocks,2):
ax.axvspan(tpb* ib, tpb*(ib+1), alpha=0.1, color='grey')
ax = axes [4,pi]
# ax.hist(1.+wMD2PFCMults[:,p*pi:p*pi+p, 0].flatten(), alpha=0.7 )#, 'tab:blue') # take a slice from context 1 #[traini, tstep, Nneur]
# ax.hist(1.+wMD2PFCMults[:,p*pi:p*pi+p, 1].flatten(), alpha=0.4 )#, 'tab:blue') # take a slice from context 1 #[traini, tstep, Nneur]
# pltu.axes_labels(ax, 'mul w values', 'freq')
ax = axes [4,0] # Need ato monitor MDpretrace in a V1 vs V2 context, but also Up and Down trials. You catch the first four trials in V1 and find up and down, the the first couple in V2, get up and down.
ax.hist(MDpreTraces[0,p*0:p*0+p].flatten(), alpha=0.7 )#, 'tab:blue') # take a slice from context 1 #[traini, tstep, Nneur]
ax.hist(MDpreTraces[0,p*1:p*1+p].flatten(), alpha=0.7 )#, 'tab:blue') # take a slice from context 1 #[traini, tstep, Nneur]
pltu.axes_labels(ax, 'MDpre trial 0', 'freq')
ax = axes [4,1] # Need ato monitor MDpretrace in a V1 vs V2 context, but also Up and Down trials. You catch the first four trials in V1 and find up and down, the the first couple in V2, get up and down.
ax.hist(MDpreTraces[1,p*0:p*0+p].flatten(), alpha=0.7 )#, 'tab:blue') # take a slice from context 1 #[traini, tstep, Nneur]
ax.hist(MDpreTraces[1,p*1:p*1+p].flatten(), alpha=0.7 )#, 'tab:blue') # take a slice from context 1 #[traini, tstep, Nneur]
pltu.axes_labels(ax, 'MDpre trial 1', 'freq')
ax = axes [4,2]
ax.hist(MDpreTraces[tpb,p*0:p*0+p].flatten(), alpha=0.7 )#, 'tab:blue') # take a slice from context 1 #[traini, tstep, Nneur]
ax.hist(MDpreTraces[tpb,p*1:p*1+p].flatten(), alpha=0.7 )#, 'tab:blue') # take a slice from context 1 #[traini, tstep, Nneur]
pltu.axes_labels(ax, 'MDpre trial tpb', 'freq')
# ax.hist(wOuts[-1,:,:].flatten(), 50, alpha=1. )#, 'tab:blue') # take a slice from context 1 #[traini, tstep, Nneur]
# pltu.axes_labels(ax, 'w outs', 'freq')
# axes[0,0].plot(wOuts[:,0,:5],'tab:red', linewidth= pltu.linewidth)
# axes[0,0].plot(wOuts[:,1,:5],'tab:red', linewidth= pltu.linewidth)
# pltu.beautify_plot(axes[0,0],x0min=False,y0min=False)
# pltu.axes_labels(axes[0,0],'Trials','wAto0(r) wAto1(b)')
# axes[0,1].plot(wOuts[:,0,config.Nsub:config.Nsub+5],'tab:red', linewidth= pltu.linewidth)
# axes[0,1].plot(wOuts[:,1,config.Nsub:config.Nsub+5],'tab:red', linewidth= pltu.linewidth)
# pltu.beautify_plot(axes[0,1],x0min=False,y0min=False)
# pltu.axes_labels(axes[0,1],'Trials','wBto0(r) wBto1(b)')
# axes[0,2].plot(wOuts[:,0,config.Nsub*2:config.Nsub*2+5],'tab:red', linewidth= pltu.linewidth)
# axes[0,2].plot(wOuts[:,1,config.Nsub*2:config.Nsub*2+5],'tab:red', linewidth= pltu.linewidth)
# pltu.beautify_plot(axes[0,2],x0min=False,y0min=False)
# pltu.axes_labels(axes[0,2],'Trials','wCto0(r) wCto1(b)')
# # pfcmd.figWeights.tight_layout()
# if config.MDlearn:
# # plot PFC2MD weights evolution
# # pfcmd.figWeights = plt.figure(
# # figsize=(pltu.twocolumnwidth,pltu.twocolumnwidth),
# # facecolor='w')
# axes[1,0].plot(wPFC2MDs[:,0,:5],'tab:red', linewidth= pltu.linewidth)
# axes[1,0].plot(wPFC2MDs[:,0,config.Nsub*2:config.Nsub*2+5],'tab:red', linewidth= pltu.linewidth)
# pltu.beautify_plot(axes[1,0],x0min=False,y0min=False)
# pltu.axes_labels(axes[1,0],'','A -> MD0(r) C (b)')
# axes[1,1].plot(wPFC2MDs[:,1,:5],'tab:red', linewidth= pltu.linewidth)
# axes[1,1].plot(wPFC2MDs[:,1,config.Nsub*2:config.Nsub*2+5],'tab:red', linewidth= pltu.linewidth)
# pltu.beautify_plot(axes[1,1],x0min=False,y0min=False)
# pltu.axes_labels(axes[1,1],'','wA->MD1(r) C->MD1(b)')
if config.reinforceReservoir:
axes[1,2].plot(wJrecs[:,1,:5],'tab:red', linewidth= pltu.linewidth)
axes[1,2].plot(wJrecs[:,-1,-5:],'tab:red', linewidth= pltu.linewidth)
pltu.beautify_plot(axes[1,2],x0min=False,y0min=False)
pltu.axes_labels(axes[1,2],'Trials','wRec1(r) wRec40(b)')
# plot MD2PFC weights evolution
# pfcmd.figWeights = plt.figure(
# figsize=(pltu.columnwidth,pltu.columnwidth),
# facecolor='w')
axes[2,0].plot(wMD2PFCs[:,:5,0],'r')
axes[2,0].plot(wMD2PFCs[:,config.Nsub*2:config.Nsub*2+5,0],'tab:red', linewidth= pltu.linewidth)
pltu.beautify_plot(axes[2,0],x0min=False,y0min=False)
pltu.axes_labels(axes[2,0],'Trials','MD 0->A (r) 0->C (b)')
axes[2,1].plot(wMD2PFCMults[:,:5,0],'tab:red', linewidth= pltu.linewidth)
axes[2,1].plot(wMD2PFCMults[:,config.Nsub*2:config.Nsub*2+5,0],'tab:red', linewidth= pltu.linewidth)
pltu.beautify_plot(axes[2,1],x0min=False,y0min=False)
pltu.axes_labels(axes[2,1],'Trials','Mw MD0toA(r) 0->C (b)')
# config.figWeights.tight_layout()
axes[3,0].plot(wMD2PFCs[:,:5,0],'tab:red', linewidth= pltu.linewidth)
axes[3,0].plot(wMD2PFCs[:,config.Nsub*2:config.Nsub*2+5,0],'tab:red', linewidth= pltu.linewidth)
pltu.beautify_plot(axes[3,0],x0min=False,y0min=False)
pltu.axes_labels(axes[3,0],'Trials','MD 1->A (r) 1->C (b)')
axes[3,1].plot(wMD2PFCMults[:,:5,0],'tab:red', linewidth= pltu.linewidth)
axes[3,1].plot(wMD2PFCMults[:,config.Nsub*2:config.Nsub*2+5,0],'tab:red', linewidth= pltu.linewidth)
pltu.beautify_plot(axes[3,1],x0min=False,y0min=False)
pltu.axes_labels(axes[3,1],'Trials','Mw MD1toA(r) 1->C (b)')
pfcmd.figWeights.tight_layout()
clip_on=False)
line_ref, = ax.plot([], [], 'o-r', lw=2, clip_on=False)
line_prediction, = ax.plot([], [], 'o-b', lw=2, clip_on=False)
time_text = ax.text(0.2,
0.78,
'',
transform=ax.transAxes,
fontsize=label_font_size)
beautify_plot(ax,
x0min=False,
y0min=False,
xticks=[],
yticks=[],
drawxaxis=False,
drawyaxis=False)
axes_labels(ax, '', '$\longleftarrow$ gravity', xpad=-20)
ax.text(0.45, 0.86, 'Acrobot', transform=fig.transFigure)
def init():
line_ref.set_data([], [])
line_prediction.set_data([], [])
time_text.set_text('')
return line_ref, line_prediction, time_text
q = q0.copy()
dq = dq0.copy()
T = 10.
t = np.arange(0, T, dt)
kernel = george.kernels.ExpSquaredKernel(.5)
def train(self,Ntrain):
MDeffect = self.MDeffect
if self.blockTrain:
Nextra = 200 # add cycles to show if block1 learning is remembered
Ntrain = Ntrain*self.Ntasks + Nextra
else:
Ntrain *= self.Ntasks
wOuts = np.zeros(shape=(Ntrain,self.Nout,self.Nneur))
if self.MDlearn:
wPFC2MDs = np.zeros(shape=(Ntrain,2,self.Nneur))
wMD2PFCs = np.zeros(shape=(Ntrain,self.Nneur,2))
wMD2PFCMults = np.zeros(shape=(Ntrain,self.Nneur,2))
MDpreTraces = np.zeros(shape=(Ntrain,self.Nneur))
# Reset the trained weights,
# earlier for iterating over MDeffect = False and then True
if self.outExternal:
self.wOut = np.random.uniform(-1,1,
size=(self.Nout,self.Nneur))/self.Nneur
self.wOut *= self.wOutMask
elif not MDeffect:
self.Jrec[-self.Nout:,:] = \
np.random.normal(size=(self.Nneur, self.Nneur))\
*self.G/np.sqrt(self.Nsub*2)
# direct connections from cue to output,
# similar to having output neurons within reservoir
if self.dirConn:
self.wDir = np.random.uniform(-1,1,
size=(self.Nout,self.Ncues))\
/self.Ncues *1.5
MSEs = np.zeros(Ntrain)
for traini in range(Ntrain):
if self.plotFigs: print('Simulating training cycle',traini)
## reduce learning rate by *10 from 100th and 200th cycle
#if traini == 100: self.learning_rate /= 10.
#elif traini == 200: self.learning_rate /= 10.
# if blockTrain,
# first half of trials is context1, second half is context2
if self.blockTrain:
taski = traini // ((Ntrain-Nextra)//self.Ntasks)
# last block is just the first context again
if traini >= Ntrain-Nextra: taski = 0
cueList = self.get_cue_list(taski)
else:
cueList = self.get_cue_list()
cues_order = self.get_cues_order(cueList)
for taski,cuei in cues_order:
cue, target = \
self.get_cue_target(taski,cuei)
cues, routs, outs, MDouts, errors = \
self.sim_cue(taski,cuei,cue,target,MDeffect=MDeffect,
train=True)
MSEs[traini] += np.mean(errors*errors)
wOuts[traini,:,:] = self.wOut
if self.plotFigs and self.outExternal:
if self.MDlearn:
wPFC2MDs[traini,:,:] = self.wPFC2MD
wMD2PFCs[traini,:,:] = self.wMD2PFC
wMD2PFCMults[traini,:,:] = self.wMD2PFCMult
MDpreTraces[traini,:] = self.MDpreTrace
self.meanAct /= Ntrain
if self.saveData:
self.fileDict['MSEs'] = MSEs
self.fileDict['wOuts'] = wOuts
if self.plotFigs:
self.fig2 = plt.figure(
figsize=(pltu.columnwidth,pltu.columnwidth),
facecolor='w')
ax2 = self.fig2.add_subplot(1,1,1)
ax2.plot(MSEs)
pltu.beautify_plot(ax2,x0min=False,y0min=False)
pltu.axes_labels(ax2,'cycle num','MSE')
self.fig2.tight_layout()
# plot output weights evolution
self.fig3 = plt.figure(
figsize=(pltu.columnwidth,pltu.columnwidth),
facecolor='w')
ax3 = self.fig3.add_subplot(2,1,1)
ax3.plot(wOuts[:,0,:5],'-,r')
ax3.plot(wOuts[:,1,:5],'-,b')
pltu.beautify_plot(ax3,x0min=False,y0min=False)
pltu.axes_labels(ax3,'cycle num','wAto0(r) wAto1(b)')
ax4 = self.fig3.add_subplot(2,1,2)
ax4.plot(wOuts[:,0,self.Nsub:self.Nsub+5],'-,r')
ax4.plot(wOuts[:,1,self.Nsub:self.Nsub+5],'-,b')
pltu.beautify_plot(ax4,x0min=False,y0min=False)
pltu.axes_labels(ax4,'cycle num','wBto0(r) wBto1(b)')
self.fig3.tight_layout()
if self.MDlearn:
# plot PFC2MD weights evolution
self.fig3 = plt.figure(
figsize=(pltu.columnwidth,pltu.columnwidth),
facecolor='w')
ax3 = self.fig3.add_subplot(2,1,1)
ax3.plot(wPFC2MDs[:,0,:5],'-,r')
ax3.plot(wPFC2MDs[:,0,self.Nsub*2:self.Nsub*2+5],'-,b')
pltu.beautify_plot(ax3,x0min=False,y0min=False)
pltu.axes_labels(ax3,'cycle num','wAtoMD0(r) wCtoMD0(b)')
ax4 = self.fig3.add_subplot(2,1,2)
ax4.plot(wPFC2MDs[:,1,:5],'-,r')
ax4.plot(wPFC2MDs[:,1,self.Nsub*2:self.Nsub*2+5],'-,b')
pltu.beautify_plot(ax4,x0min=False,y0min=False)
pltu.axes_labels(ax4,'cycle num','wAtoMD1(r) wCtoMD1(b)')
self.fig3.tight_layout()
# plot MD2PFC weights evolution
self.fig3 = plt.figure(
figsize=(pltu.columnwidth,pltu.columnwidth),
facecolor='w')
ax3 = self.fig3.add_subplot(2,1,1)
ax3.plot(wMD2PFCs[:,:5,0],'-,r')
ax3.plot(wMD2PFCs[:,self.Nsub*2:self.Nsub*2+5,0],'-,b')
pltu.beautify_plot(ax3,x0min=False,y0min=False)
pltu.axes_labels(ax3,'cycle num','wMD0toA(r) wMD0toC(b)')
ax4 = self.fig3.add_subplot(2,1,2)
ax4.plot(wMD2PFCMults[:,:5,0],'-,r')
ax4.plot(wMD2PFCMults[:,self.Nsub*2:self.Nsub*2+5,0],'-,b')
pltu.beautify_plot(ax4,x0min=False,y0min=False)
pltu.axes_labels(ax4,'cycle num','MwMD0toA(r) MwMD0toC(b)')
self.fig3.tight_layout()
# plot PFC to MD pre Traces
self.fig3 = plt.figure(
figsize=(pltu.columnwidth,pltu.columnwidth),
facecolor='w')
ax3 = self.fig3.add_subplot(1,1,1)
ax3.plot(MDpreTraces[:,:5],'-,r')
ax3.plot(MDpreTraces[:,self.Nsub*2:self.Nsub*2+5],'-,b')
pltu.beautify_plot(ax3,x0min=False,y0min=False)
pltu.axes_labels(ax3,'cycle num','cueApre(r) cueCpre(b)')
self.fig3.tight_layout()
def plot_column(self,fig,cues,routs,MDouts,outs,ploti=0):
print('Plotting ...')
cols=4
if ploti==0:
yticks = (0,1)
ylabels=('Cues','PFC for cueA','PFC for cueB',
'PFC for cueC','PFC for cueD','PFC for rest',
'MD 1,2','Output 1,2')
else:
yticks = ()
ylabels=('','','','','','','','')
ax = fig.add_subplot(8,cols,1+ploti)
ax.plot(cues,linewidth=pltu.plot_linewidth)
ax.set_ylim([-0.1,1.1])
pltu.beautify_plot(ax,x0min=False,y0min=False,
xticks=(),yticks=yticks)
pltu.axes_labels(ax,'',ylabels[0])
ax = fig.add_subplot(8,cols,cols+1+ploti)
ax.plot(routs[:,:10],linewidth=pltu.plot_linewidth)
ax.set_ylim([-0.1,1.1])
pltu.beautify_plot(ax,x0min=False,y0min=False,
xticks=(),yticks=yticks)
pltu.axes_labels(ax,'',ylabels[1])
ax = fig.add_subplot(8,cols,cols*2+1+ploti)
ax.plot(routs[:,self.Nsub:self.Nsub+10],
linewidth=pltu.plot_linewidth)
ax.set_ylim([-0.1,1.1])
pltu.beautify_plot(ax,x0min=False,y0min=False,
xticks=(),yticks=yticks)
pltu.axes_labels(ax,'',ylabels[2])
if self.Ncues > 2:
ax = fig.add_subplot(8,cols,cols*3+1+ploti)
ax.plot(routs[:,self.Nsub*2:self.Nsub*2+10],
linewidth=pltu.plot_linewidth)
ax.set_ylim([-0.1,1.1])
pltu.beautify_plot(ax,x0min=False,y0min=False,
xticks=(),yticks=yticks)
pltu.axes_labels(ax,'',ylabels[3])
ax = fig.add_subplot(8,cols,cols*4+1+ploti)
ax.plot(routs[:,self.Nsub*3:self.Nsub*3+10],
linewidth=pltu.plot_linewidth)
ax.set_ylim([-0.1,1.1])
pltu.beautify_plot(ax,x0min=False,y0min=False,
xticks=(),yticks=yticks)
pltu.axes_labels(ax,'',ylabels[4])
ax = fig.add_subplot(8,cols,cols*5+1+ploti)
ax.plot(routs[:,self.Nsub*4:self.Nsub*4+10],
linewidth=pltu.plot_linewidth)
ax.set_ylim([-0.1,1.1])
pltu.beautify_plot(ax,x0min=False,y0min=False,
xticks=(),yticks=yticks)
pltu.axes_labels(ax,'',ylabels[5])
ax = fig.add_subplot(8,cols,cols*6+1+ploti)
ax.plot(MDouts,linewidth=pltu.plot_linewidth)
ax.set_ylim([-0.1,1.1])
pltu.beautify_plot(ax,x0min=False,y0min=False,
xticks=(),yticks=yticks)
pltu.axes_labels(ax,'',ylabels[6])
ax = fig.add_subplot(8,cols,cols*7+1+ploti)
ax.plot(outs,linewidth=pltu.plot_linewidth)
ax.set_ylim([-0.1,1.1])
pltu.beautify_plot(ax,x0min=False,y0min=False,
xticks=[0,self.tsteps],yticks=yticks)
pltu.axes_labels(ax,'time (ms)',ylabels[7])
fig.tight_layout()
if self.saveData:
d = {}
# 1st column of all matrices is number of time steps
# 2nd column is number of neurons / units
d['cues'] = cues # tsteps x 4
d['routs'] = routs # tsteps x 1000
d['MDouts'] = MDouts # tsteps x 2
d['outs'] = outs # tsteps x 2
savemat('simData'+str(ploti), d)
return ax
def train(self, Ntrain):
if self.blockTrain:
Nextra = 200 # add cycles to show if block1 learning is remembered
Ntrain = Ntrain * self.Ntasks + Nextra
else:
Ntrain *= self.Ntasks
wOuts = np.zeros(shape=(Ntrain, self.Nout, self.Nneur))
# Reset the trained weights,
# earlier for iterating over MDeffect = False and then True
self.wOut = np.random.uniform(
-1, 1, size=(self.Nout, self.Nneur)) / self.Nneur
MSEs = np.zeros(Ntrain)
for traini in range(Ntrain):
if self.plotFigs: print('Simulating training cycle', traini)
## reduce learning rate by *10 from 100th and 200th cycle
#if traini == 100: self.learning_rate /= 10.
#elif traini == 200: self.learning_rate /= 10.
# if blockTrain,
# first half of trials is context1, second half is context2
if self.blockTrain:
taski = traini // ((Ntrain - Nextra) // self.Ntasks)
# last block is just the first context again
if traini >= Ntrain - Nextra: taski = 0
cueList = self.get_cue_list(taski)
else:
cueList = self.get_cue_list()
cues_order = self.get_cues_order(cueList)
for taski, cuei in cues_order:
cue, target = \
self.get_cue_target(taski,cuei)
cues, routs, outs, MDouts, errors = \
self.sim_cue(taski,cuei,cue,target,train=True)
MSEs[traini] += np.mean(errors * errors)
wOuts[traini, :, :] = self.wOut
self.meanAct /= Ntrain
if self.saveData:
self.fileDict['MSEs'] = MSEs
self.fileDict['wOuts'] = wOuts
if self.plotFigs:
self.fig2 = plt.figure(figsize=(pltu.columnwidth,
pltu.columnwidth),
facecolor='w')
ax2 = self.fig2.add_subplot(1, 1, 1)
ax2.plot(MSEs)
pltu.beautify_plot(ax2, x0min=False, y0min=False)
pltu.axes_labels(ax2, 'cycle num', 'MSE')
self.fig2.tight_layout()
# plot output weights evolution
self.fig3 = plt.figure(figsize=(pltu.columnwidth,
pltu.columnwidth),
facecolor='w')
ax3 = self.fig3.add_subplot(2, 1, 1)
ax3.plot(wOuts[:, 0, :5], '-,r')
ax3.plot(wOuts[:, 1, :5], '-,b')
pltu.beautify_plot(ax3, x0min=False, y0min=False)
pltu.axes_labels(ax3, 'cycle num', 'wAto0(r) wAto1(b)')
ax4 = self.fig3.add_subplot(2, 1, 2)
ax4.plot(wOuts[:, 0, self.Nsub:self.Nsub + 5], '-,r')
ax4.plot(wOuts[:, 1, self.Nsub:self.Nsub + 5], '-,b')
pltu.beautify_plot(ax4, x0min=False, y0min=False)
pltu.axes_labels(ax4, 'cycle num', 'wBto0(r) wBto1(b)')
self.fig3.tight_layout()