def perform_analysis(self):
self.datastore.print_content()
dsv = analysis_data_structure_parameter_filter_query(self.datastore,identifier='PerNeuronValue')
for sheet in self.datastore.sheets():
# Get PerNeuronValue ASD and make sure they are all associated with the same stimulus are
# do not differ in any ASD parameters except the stimulus
dsv = select_result_sheet_query(dsv,sheet)
if ads_is_empty(dsv): break
assert equal_ads_except(dsv,['stimulus_id'])
assert ads_with_equal_stimulus_type(dsv,not_None=True)
self.pnvs = dsv.get_analysis_result(sheet_name=sheet)
# get stimuli
st = [StimulusID(s.stimulus_id) for s in self.pnvs]
d = colapse_to_dictionary([z.values for z in self.pnvs],st,self.parameters.parameter_name)
result_dict = {}
for k in d.keys():
keys,values = d[k]
y = []
x = []
for v,p in zip(values,keys):
y.append(v)
x.append(numpy.zeros(numpy.shape(v))+p)
pref,sel = circ_mean(numpy.array(x),weights=numpy.array(y),axis=0,low=0,high=st[0].periods[self.parameters.parameter_name],normalize=True)
logger.debug('Adding PerNeuronValue to datastore')
self.datastore.full_datastore.add_analysis_result(PerNeuronValue(pref,st[0].units[self.parameters.parameter_name],value_name = self.parameters.parameter_name + ' preference',sheet_name=sheet,tags=self.tags,period=st[0].periods[self.parameters.parameter_name],analysis_algorithm=self.__class__.__name__,stimulus_id=str(k)))
self.datastore.full_datastore.add_analysis_result(PerNeuronValue(sel,st[0].units[self.parameters.parameter_name],value_name = self.parameters.parameter_name + ' selectivity',sheet_name=sheet,tags=self.tags,period=1.0,analysis_algorithm=self.__class__.__name__,stimulus_id=str(k)))
def __init__(self, network, source, target, parameters,name):
MozaikComponent.__init__(self,network,parameters)
self.name = name
self.source = source
self.target = target
weights = []
for (i,neuron1) in enumerate(self.target.pop.all()):
for (j,neuron2) in enumerate(self.source.pop.all()):
or_dist = circular_dist(self.target.get_neuron_annotation(i,'LGNAfferentOrientation'),self.source.get_neuron_annotation(j,'LGNAfferentOrientation'),numpy.pi) / (numpy.pi/2)
if self.parameters.specific_arborization.target_synapses == 'excitatory':
phase_dist = circular_dist(self.target.get_neuron_annotation(i,'LGNAfferentPhase'),self.source.get_neuron_annotation(j,'LGNAfferentPhase'),2*numpy.pi) / numpy.pi
elif self.parameters.specific_arborization.target_synapses == 'inhibitory':
phase_dist = (numpy.pi - circular_dist(self.target.get_neuron_annotation(i,'LGNAfferentPhase'),self.source.get_neuron_annotation(j,'LGNAfferentPhase'),2*numpy.pi)) / numpy.pi
else:
logger.error('Unknown type of synapse!')
return
or_gauss = normal_function(or_dist,mean=0,sigma=self.parameters.or_sigma)
phase_gauss = normal_function(phase_dist,mean=0,sigma=self.parameters.or_sigma)
w = phase_gauss*or_gauss*or_gauss
weights.append((j,i,w,self.parameters.propagation_constant))
we = numpy.zeros((len(self.source.pop),len(self.target.pop)))
pnv_source = []
pnv_target = []
for (i,neuron1) in enumerate(self.target.pop.all()):
pnv_target.append(self.target.get_neuron_annotation(i,'LGNAfferentOrientation'))
for (i,neuron1) in enumerate(self.source.pop.all()):
pnv_source.append(self.source.get_neuron_annotation(i,'LGNAfferentOrientation'))
for (j,i,w,xsj) in weights:
we[j,i] = w
(angle,mag) = circ_mean(numpy.array([numpy.array(pnv_target).T for i in xrange(0,numpy.shape(we)[0])]),weights=numpy.array(we),low=0,high=numpy.pi,axis=1)
if self.parameters.probabilistic:
SpecificProbabilisticArborization(network, self.source, self.target, weights,self.parameters.specific_arborization,self.name).connect()
else:
SpecificArborization(network, self.source, self.target, weights,self.parameters.specific_arborization,self.name).connect()
def ploter(self,idx,gs,params):
sx = self.connecting_neurons_positions[idx][0]
sy = self.connecting_neurons_positions[idx][1]
tx = self.connected_neuron_position[idx][0]
ty = self.connected_neuron_position[idx][1]
if not self.parameters.reversed:
w = self.connections[idx].weights[self.parameters.neuron,:]
else:
w = self.connections[idx].weights[:,self.parameters.neuron]
# pick the right PerNeuronValue to show
pnv = []
if self.pnvs!=None:
for p in self.pnvs:
if not self.parameters.reversed and p.sheet_name == self.connections[idx].target_name:
pnv.append(p)
if self.parameters.reversed and p.sheet_name == self.connections[idx].source_name:
pnv.append(p)
if len(pnv) > 1:
raise ValueError('ERROR: too many matching PerNeuronValue ADSs')
else:
pnv = pnv[0]
if len(pnv.values) != len(w):
raise ValueError('ERROR: length of colors does not match length of weights \[%d \!\= %d\]. Ignoring colors!' % (len(pnv.values),len(w)))
if pnv != None:
from mozaik.tools.circ_stat import circ_mean
(angle,mag) = circ_mean(numpy.array(pnv.values),weights=w,high=pnv.period)
params.setdefault("title",str(self.connections[idx].name) + "| Weighted mean: " + str(angle))
params.setdefault("colorbar_label",pnv.value_name)
params.setdefault("colorbar",True)
if self.connections[idx].source_name == self.connections[idx].target_name:
ConnectionPlot(sx,sy,tx,ty,w,colors=pnv.values,line=False,period=pnv.period,**params)(gs)
else:
ConnectionPlot(sx,sy,numpy.min(sx)*1.2,numpy.min(sy)*1.2,w,colors=pnv.values,period=pnv.period,line=False,**params)(gs)
else:
params.setdefault("title",self.connections[idx].name)
if self.connections[idx].source_name == self.connections[idx].target_name:
ConnectionPlot(sx,sy,tx,ty,w,line=False,**params)(gs)
else:
ConnectionPlot(sx,sy,numpy.min(sx)*1.2,numpy.min(sy)*1.2,w,line=False,**params)(gs)
def _ploter(self, idx, gs):
#ix = numpy.flatnonzero(self.connections[idx].weights[:,1]==index)
#x = self.proj.pre.positions[0][self.connections[idx].weights[ix,0]]
#y = self.proj.pre.positions[1][self.connections[idx].weights[ix,0]]
#w = weights[idx,2]
tx = self.connected_neuron_position[idx][0]
ty = self.connected_neuron_position[idx][1]
if not self.parameters.reversed:
index = self.datastore.get_sheet_indexes(self.connections[idx].source_name,self.parameters.neuron)
ix = numpy.flatnonzero(numpy.array(self.connections[idx].weights)[:,1]==index)
else:
index = self.datastore.get_sheet_indexes(self.connections[idx].target_name,self.parameters.neuron)
ix = numpy.flatnonzero(numpy.array(self.connections[idx].weights)[:,0]==index)
sx = self.connecting_neurons_positions[idx][0][ix]
sy = self.connecting_neurons_positions[idx][1][ix]
w = numpy.array(self.connections[idx].weights)[ix,2]
d = numpy.array(self.connections[idx].delays)[ix,2]
assert numpy.shape(w) == numpy.shape(d)
# pick the right PerNeuronValue to show
pnv = []
if self.pnvs != None:
for p in self.pnvs:
if not self.parameters.reversed and p.sheet_name == self.connections[idx].target_name:
pnv.append(p)
if self.parameters.reversed and p.sheet_name == self.connections[idx].source_name:
pnv.append(p)
if len(pnv) > 1:
raise ValueError('ERROR: too many matching PerNeuronValue ADSs')
elif len(pnv) != 0:
pnv = pnv[0]
if len(pnv.values) != len(w):
raise ValueError('ERROR: length of colors does not match length of weights \[%d \!\= %d\]. Ignoring colors!' % (len(pnv.values), len(w)))
gss = gridspec.GridSpecFromSubplotSpec(2, 1, subplot_spec=gs)
# Plot the weights
gs = gss[0,0]
params = {}
if pnv != []:
from mozaik.tools.circ_stat import circ_mean
(angle, mag) = circ_mean(numpy.array(pnv.values),
weights=w,
high=pnv.period)
params["title"] = 'Weights: '+ str(self.connections[idx].proj_name) + "| Weighted mean: " + str(angle)
params["colorbar_label"] = pnv.value_name
params["colorbar"] = True
if self.connections[idx].source_name == self.connections[idx].target_name:
params["line"] = False
plots = [("ConnectionsPlot",ConnectionPlot(sx, sy, tx, ty, w,colors=pnv.values,period=pnv.period),gs,params)]
else:
params["line"] = True
plots = [("ConnectionsPlot",ConnectionPlot(sx, sy, numpy.min(sx)*1.2, numpy.min(sy)*1.2, w,colors=pnv.values,period=pnv.period),gs,params)]
else:
params["title"] = 'Weights: '+ self.connections[idx].proj_name
if self.connections[idx].source_name == self.connections[idx].target_name:
params["line"] = False
plots = [("ConnectionsPlot",ConnectionPlot(sx, sy, tx, ty, w,colors=pnv.values,period=pnv.period),gs,params)]
else:
params["line"] = True
plots = [("ConnectionsPlot",ConnectionPlot(sx, sy, numpy.min(sx)*1.2, numpy.min(sy)*1.2,w),gs,params)]
# Plot the delays
gs = gss[1,0]
params = {}
params["title"] = 'Delays: '+ self.connections[idx].proj_name
if self.connections[idx].source_name == self.connections[idx].target_name:
params["line"] = False
plots.append(("DelaysPlot",ConnectionPlot(sx, sy, tx, ty, (numpy.zeros(w.shape)+0.3)*(w!=0),colors=d),gs,params))
else:
params["line"] = True
plots.append(("DelaysPlot",ConnectionPlot(sx, sy, numpy.min(sx)*1.2, numpy.min(sy)*1.2,(numpy.zeros(w.shape)+0.3)*(w!=0),colors=d),gs,params))
return plots
