def forward_LQLEP( stimulus, all_spikes, model, indices, vardict=u2c_parameterization()):
print
print 'Preparing forward LQLEP model.'
sys.stdout.flush()
vardict = Poisson_LL( **RGC_LE( **subunit_LQ( **vardict)))
variables = ['stimulus','all_spikes','u','V2','sv1','T','c','uc']
knowns = model
knowns.update({'T':retina.place_cells( cones, cones, shapes )})
unknowns = {'stimulus':stimulus, 'all_spikes':all_spikes}
outputs = {'LL': vardict['loglikelihood'], 'rates': vardict['rgc_out']}
# 'theta': vardict['theta']}
NRGC = len( all_spikes )
return make_global_objective(unknowns,knowns,vardict,variables,outputs,indices,NRGC)
def global_objective( unknowns, knowns, vardict, run, indices ):
print
print 'Preparing objective for unknowns: ', [(n,v.shape) for n,v in unknowns.items()]
sys.stdout.flush()
variables = ['u','V2','sv1','sparse_STA','sparse_STC','V1','N_spikes','T','cov']
run.update({'u':init_u, 'V2':init_V2, 'sv1':init_sv1(run['rgc_type']),
'T':retina.place_cells( cones, cones, shapes )})
default( knowns, run, variables )
outputs = {}
update_dict( outputs, 'f', vardict ,
['LNP', 'LQLEP_wPrior', 'LQLEP_positiveV1', 'LQLEP_positive_u'] )
update_dict( outputs , 'LL' , vardict , ['LNP', 'LQLEP'] )
update_dict( outputs , 'barrier' , vardict ,
['barrier','barrier_positiveV1','barrier_positive_u'] )
update_dict( outputs , 'nonlinearity', vardict , ['nonlinearity'] )
if unknowns.has_key('sv1'):
print
print 'sv1 shape', unknowns['sv1'].shape
print
NRGC = len( run['sparse_STA'] )
return make_global_objective(unknowns,knowns,vardict,variables,outputs,indices,NRGC)