def load_dic(self, *filt, **kw):
d = {}
if kw.get('keys_iterator'):
for keys in kw['keys_iterator']:
print keys
storage = misc.dict_recursive_get(self, keys)
val = storage.load_data()
d = misc.dict_recursive_add(d, keys, val)
return d
for keys, storage in misc.dict_iter(self):
if filt == ():
pass
else:
a = False
i = 0
for key in keys:
if key not in filt and i not in kw.get('star', []):
a = True
i += 1
if key == 'mean_rates':
pass
if i == len(keys):
break
if a:
continue
val = storage.load_data()
d = misc.dict_recursive_add(d, keys, val)
return d
def load_dic(self, *filt, **kw):
d={}
if kw.get('keys_iterator'):
for keys in kw['keys_iterator']:
print keys
storage=misc.dict_recursive_get(self, keys)
val=storage.load_data()
d=misc.dict_recursive_add(d, keys, val)
return d
for keys, storage in misc.dict_iter(self):
if filt==():
pass
else:
a=False
i=0
for key in keys:
if key not in filt and i not in kw.get('star', []):
a=True
i+=1
if key=='mean_rates':
pass
if i==len(keys):
break
if a:
continue
val=storage.load_data()
d=misc.dict_recursive_add(d, keys, val)
return d
def set_optimization_val(data, nets):
for key in sorted(nets.keys()):
net = nets[key]
inp = net.get_single_unit_input()
x = 'node.' + inp + '.rate' #kwargs.get('x', ['node.CFp.rate'])
dic = misc.dict_recursive_add({}, x.split('.'), data.xopt[-1][0])
net.update_dic_rep(dic)
def cmp_mean_rate_diff(d, models, parings, x, _set='set_0'):
for model in models:
y=[]
y_std=[]
for net0, net1 in parings:
d0=d[net0][_set]
d1=d[net1][_set]
v0=d0[model]['mean_rate_slices']
v1=d1[model]['mean_rate_slices']
y.append(numpy.mean(numpy.abs(v0.y_raw_data-v1.y_raw_data)))
y_std.append(numpy.std(numpy.abs(v0.y_raw_data-v1.y_raw_data)))
dd={'y':numpy.array(y),
'y_std':numpy.array(y_std),
'x':numpy.array(x)}
obj=Data_generic(**dd)
d=misc.dict_recursive_add(d, ['Difference',
model,
'mean_rate_diff'], obj)
pp(d)
return d
def generate_plot_data(d,
models,
attrs,
exclude=['all', 'striatum', 'GP-ST-SN', 'Normal']):
out = {}
labels = []
keys = d.keys()
for k in sorted(keys):
if k in exclude:
continue
labels.append(k)
for model in models:
l = []
for attr in attrs:
l.append([])
for k in sorted(keys):
if k in exclude:
continue
l[-1].append(misc.dict_recursive_get(d, [k, model, attr]))
lsum = [[a] * len(l[0]) for a in numpy.sum(l, axis=1)]
obj = Data_bar(**{
'y': numpy.array(l) #/numpy.array(lsum)
})
out = misc.dict_recursive_add(out, [model], obj)
return out, labels
def generate_plot_data(d, models, attrs, exclude=['all',
'striatum',
'GP-ST-SN',
'Normal']):
out={}
labels=[]
keys=d.keys()
for k in sorted(keys):
if k in exclude:
continue
labels.append(k)
for model in models:
l=[]
for attr in attrs:
l.append([])
for k in sorted(keys):
if k in exclude:
continue
l[-1].append(misc.dict_recursive_get(d, [k, model, attr]))
lsum=[[a]*len(l[0]) for a in numpy.sum(l, axis=1)]
obj=Data_bar(**{'y':numpy.array(l)#/numpy.array(lsum)
})
out=misc.dict_recursive_add(out, [model], obj)
return out, labels
def show_plot(axs,
name,
d,
models=['M1', 'M2', 'FS', 'GA', 'GI', 'ST', 'SN'],
**k):
dd = {}
by_sets = k.pop('by_sets', False)
for keys, val in misc.dict_iter(d):
if keys[-1] != name:
continue
if by_sets and keys[0][0:3] != 'set':
continue
first_keys = keys[:-2]
if type(first_keys) == str:
first_keys = [first_keys]
new_keys = ['_'.join(first_keys)] + keys[-2:]
dd = misc.dict_recursive_add(dd, new_keys, val)
d = dd
# if k.get('fig_and_axes', False):
# else:
# fig, axs=ps.get_figure(n_rows=len(models), n_cols=1, w=1000.0, h=800.0,
# fontsize=k.get('fontsize',10))
labels = k.pop('labels', sorted(d.keys()))
# colors=misc.make_N_colors('Paired', max(len(labels), 6))
colors = k.pop('colors', misc.make_N_colors('jet', max(len(labels), 1)))
linestyles = ['-'] * len(labels)
j = 0
nets = k.get('nets', sorted(d.keys()))
for key in nets:
v = d[key]
# axs[0].set_title(k)
for i, model in enumerate(models):
kk = {
'label': model + ' ' + labels[j],
'linestyle': linestyles[j],
'color': colors[j]
}
if 'win' in k.keys():
kk['win'] = k['win']
if 't_start' in k.keys():
kk['t_start'] = k['t_start']
if 't_stop' in k.keys():
kk['t_stop'] = k['t_stop']
print model
if model in v.keys():
v[model][name].plot(ax=axs[i], **kk)
j += 1
for ax in axs:
ax.legend()
def process_exp(exp, tr):
out = {}
put_at = {'STN': 3, 'TA': 0, 'TI': 1, 'all': 2}
for keys, val in misc.dict_iter(exp):
if keys[-1] == 'CV':
continue
keys2 = [tr[keys[1]], tr[keys[2]], 'mean_rates']
if not misc.dict_haskey(out, keys2):
a = numpy.zeros(4)
a[put_at[keys[0]]] = val
misc.dict_recursive_add(out, keys2, a)
else:
out[tr[keys[1]]][tr[keys[2]]]['mean_rates'][put_at[keys[0]]] = val
pp(out)
return out
def extract_data(d, nets, models, attrs):
out={}
for keys, val in misc.dict_iter(d):
if keys[-1]=='phases_diff_with_cohere':
v=numpy.mean(val.y_val, axis=0)
if keys[-1]=='mean_coherence':
v=val.y[2:20]
out=misc.dict_recursive_add(out, keys, v)
return out
def process_exp(exp, tr):
out={}
put_at={'STN':3,
'TA':0,
'TI':1,
'all':2}
for keys, val in misc.dict_iter(exp):
if keys[-1]=='CV':
continue
keys2=[tr[keys[1]],tr[keys[2]], 'mean_rates']
if not misc.dict_haskey(out, keys2):
a=numpy.zeros(4)
a[put_at[keys[0]]]=val
misc.dict_recursive_add(out, keys2,a )
else:
out[tr[keys[1]]] [tr[keys[2]]] ['mean_rates'][put_at[keys[0]]]=val
pp(out)
return out
def extract_data(d, nets, models, attrs):
out = {}
for keys, val in misc.dict_iter(d):
if keys[-1] == 'phases_diff_with_cohere':
v = numpy.mean(val.y_val, axis=0)
if keys[-1] == 'mean_coherence':
v = val.y[2:20]
out = misc.dict_recursive_add(out, keys, v)
return out
def compute_performance(d, nets, models, attrs):
results={}
for run in d.keys():
for model in models:
for attr in attrs:
keys1=[run, 'Net_0', model, attr]
keys2=[run, 'Net_1', model, attr]
keys3=[run, model, attr]
v1=misc.dict_recursive_get(d, keys1)
v2=misc.dict_recursive_get(d, keys2)
v=(v1-v2)/v1
v[numpy.isnan(v)]=0
v=numpy.mean((v)**2)
results=misc.dict_recursive_add(results, keys3, v)
d0={}
for run in results.keys():
for model in models:
for attr in attrs:
keys0=['Normal', model, attr]
v1=misc.dict_recursive_get(results, keys0)
d0=misc.dict_recursive_add(d0, keys0, v1)
for run in results.keys():
for model in models:
for attr in attrs:
keys0=['Normal', model, attr]
keys3=[run, model, attr]
v0=misc.dict_recursive_get(d0, keys0)
v1=misc.dict_recursive_get(results, keys3)
results=misc.dict_recursive_add(results, keys3, v1/v0)
return results
def compute_performance(d, nets, models, attrs):
results = {}
for run in d.keys():
for model in models:
for attr in attrs:
keys1 = [run, 'Net_0', model, attr]
keys2 = [run, 'Net_1', model, attr]
keys3 = [run, model, attr]
v1 = misc.dict_recursive_get(d, keys1)
v2 = misc.dict_recursive_get(d, keys2)
v = (v1 - v2) / v1
v[numpy.isnan(v)] = 0
v = numpy.mean((v)**2)
results = misc.dict_recursive_add(results, keys3, v)
d0 = {}
for run in results.keys():
for model in models:
for attr in attrs:
keys0 = ['Normal', model, attr]
v1 = misc.dict_recursive_get(results, keys0)
d0 = misc.dict_recursive_add(d0, keys0, v1)
for run in results.keys():
for model in models:
for attr in attrs:
keys0 = ['Normal', model, attr]
keys3 = [run, model, attr]
v0 = misc.dict_recursive_get(d0, keys0)
v1 = misc.dict_recursive_get(results, keys3)
results = misc.dict_recursive_add(results, keys3, v1 / v0)
return results
def postprocess(d):
out = {}
ylabels0 = sorted(d.keys())
xlabels = sorted(d[ylabels0[0]]['Net_0'].keys())
i, j = 0, 0
for tp in ['beta', 'sw']:
ylabels = []
for key in ylabels0:
if key[0] != tp[0] or len(d[key].keys()) < 2:
continue
ylabels.append(key)
j += 1
# if j==396:
# pass
for net in d[key].keys():
mrs = []
for model in xlabels:
obj = d[key][net][model]['spike_statistic']
mrs.append(obj.rates['mean'])
keys = [tp, net, 'mean_rates']
if not misc.dict_haskey(out, keys):
misc.dict_recursive_add(out, keys, [mrs])
if net == 'Net_0': i += 1
else:
out[tp][net]['mean_rates'].append(mrs)
if net == 'Net_0': i += 1
if not i == j:
print d[key].keys(), i, j
raise
# print j,i
for net in d[key].keys():
keys = [tp, net, 'xlabels']
misc.dict_recursive_add(out, keys, xlabels)
keys = [tp, net, 'ylabels']
misc.dict_recursive_add(out, keys, ylabels)
pp(out)
for tp in out.keys():
for net in out[tp].keys():
out[tp][net]['mean_rates'] = numpy.array(
out[tp][net]['mean_rates'])
return out
def load_dic(self, *filt):
d={}
for keys, storage in misc.dict_iter(self):
if filt==():
pass
else:
a=False
i=0
for key in keys:
if key not in filt:
a=True
i+=1
if i==3:
break
if a:
continue
val=storage.load_data()
d=misc.dict_recursive_add(d, keys, val)
return d
def load_dic(self, *filt):
d = {}
for keys, storage in misc.dict_iter(self):
if filt == ():
pass
else:
a = False
i = 0
for key in keys:
if key not in filt:
a = True
i += 1
if i == 3:
break
if a:
continue
val = storage.load_data()
d = misc.dict_recursive_add(d, keys, val)
return d
def postprocess(d):
out={}
ylabels0=sorted(d.keys())
xlabels=sorted(d[ylabels0[0]]['Net_0'].keys())
i,j=0,0
for tp in ['beta','sw']:
ylabels=[]
for key in ylabels0:
if key[0]!=tp[0] or len(d[key].keys())<2:
continue
ylabels.append(key)
j+=1
# if j==396:
# pass
for net in d[key].keys():
mrs=[]
for model in xlabels:
obj=d[key][net][model]['spike_statistic']
mrs.append(obj.rates['mean'])
keys=[tp, net, 'mean_rates']
if not misc.dict_haskey(out, keys):
misc.dict_recursive_add(out, keys, [mrs])
if net=='Net_0': i+=1
else:
out[tp][net]['mean_rates'].append(mrs)
if net=='Net_0':i+=1
if not i==j:
print d[key].keys(), i,j
raise
# print j,i
for net in d[key].keys():
keys=[tp, net, 'xlabels']
misc.dict_recursive_add(out, keys, xlabels)
keys=[tp, net, 'ylabels']
misc.dict_recursive_add(out, keys, ylabels)
pp(out)
for tp in out.keys():
for net in out[tp].keys():
out[tp][net]['mean_rates']=numpy.array(out[tp][net]['mean_rates'])
return out
def add_storage(self, keys):
val = self.Factory_storage()
self.dic = misc.dict_recursive_add(self.dic, keys, val)
def add_storage(self, keys):
val=self.Factory_storage()
self.dic=misc.dict_recursive_add(self.dic, keys, val)