def get_weights(self):
conn_list_fn = self.params['conn_list_ee_fn_base'] + '0.dat'
self.conn_mat, self.delays = utils.convert_connlist_to_matrix(conn_list_fn, self.params['n_exc'], self.params['n_exc'])
self.w_in_sum = np.zeros(self.params['n_exc'])
n_cells = self.params['n_exc']
for i in xrange(n_cells):
self.w_in_sum[i] = self.conn_mat[:, i].sum()
def plot_conductance_composition(self):
self.load_nspikes()
if self.no_spikes:
return
self.load_input_spikes()
self.g_exc_noise = self.params['w_exc_noise'] * self.params['tau_syn_exc'] * self.params['f_exc_noise']
self.g_inh_noise = self.params['w_inh_noise'] * self.params['tau_syn_inh'] * self.params['f_inh_noise']
conn_list_ee = self.params['merged_conn_list_ee']
conn_list_ei = self.params['merged_conn_list_ei']
conn_list_ie = self.params['merged_conn_list_ie']
conn_list_ii = self.params['merged_conn_list_ii']
if not os.path.exists(conn_list_ee):
os.system("python merge_connlists.py")
print 'Getting connection matrices ...'
w_ee, delays_ee = utils.convert_connlist_to_matrix(conn_list_ee, self.params['n_exc'], self.params['n_exc'])
w_ei, delays_ei = utils.convert_connlist_to_matrix(conn_list_ei, self.params['n_exc'], self.params['n_inh'])
w_ie, delays_ie = utils.convert_connlist_to_matrix(conn_list_ie, self.params['n_inh'], self.params['n_exc'])
w_ii, delays_ii = utils.convert_connlist_to_matrix(conn_list_ii, self.params['n_inh'], self.params['n_inh'])
c_ee = self.get_cond_matrix(self.nspikes_exc, w_ee, 'ee')
c_ei = self.get_cond_matrix(self.nspikes_exc, w_ei, 'ei')
c_ie = self.get_cond_matrix(self.nspikes_inh, w_ie, 'ie')
c_ii = self.get_cond_matrix(self.nspikes_inh, w_ii, 'ii')
self.create_fig()
self.n_fig_x, self.n_fig_y = 1, 1
color_net_exc = 'r'
color_net_inh = 'b'
color_stim = 'g'
color_noise = 'w'
ax = self.fig.add_subplot(111)
lw = 4
width=.5
for gid in xrange(self.params['n_exc']):
x = self.nspikes_exc[gid]
y0 = self.g_stim[gid]
y1 = c_ee[:, gid].sum()
y3 = c_ie[:, gid].sum()
ax.bar(x, y0, width=width, color=color_stim)
ax.bar(x, y1, width=width, bottom=y0, color=color_net_exc)
# ax.bar(x, self.g_exc_noise, width=width, bottom=y1+y0, color=color_noise)
ax.bar(x, -y3, width=width, color=color_net_inh)
def print_delays(self):
fn = self.params['merged_conn_list_ee']
if not os.path.exists(fn):
print 'File: %s not found\nCalling merge_connlists.py now...\n'
os.system('python merge_connlists.py')
conn_mat, delays = utils.convert_connlist_to_matrix(fn, self.params['n_exc'], self.params['n_exc'])
for i_, src in enumerate(self.gids_to_plot):
for j_, tgt in enumerate(self.gids_to_plot):
if src != tgt:
t1 = self.time_of_max_stim[src]
t2 = self.time_of_max_stim[tgt]
w, delay = conn_mat[src, tgt], delays[src, tgt]
if t1 < t2:
optimal_tau_prediction = (t2 - t1) / self.params['t_stimulus']
optimal_delay_scale = delay / (self.params['delay_scale'] * (t2 - t1))
print '%d (%d)\t->\t%d (%d)\t: dt_max_stim = %.1f\toptimal_tau_prediction = %.1f\toptimal_delay_scale = %.1f' % (src, t1, tgt, t2, t2 - t1, optimal_tau_prediction, optimal_delay_scale)
def load_connection_matrices(self, conn_type):
"""
deprecated - should not be used because of unnecessary memory consumption
use load_connection_list instead to get sources / targets
"""
if conn_type == 'ee':
n_src, n_tgt = self.params['n_exc'], self.params['n_exc']
loaded = self.conn_mat_loaded[0]
elif conn_type == 'ei':
n_src, n_tgt = self.params['n_exc'], self.params['n_inh']
loaded = self.conn_mat_loaded[1]
elif conn_type == 'ie':
n_src, n_tgt = self.params['n_inh'], self.params['n_exc']
loaded = self.conn_mat_loaded[2]
elif conn_type == 'ii':
n_src, n_tgt = self.params['n_inh'], self.params['n_inh']
loaded = self.conn_mat_loaded[3]
if loaded:
return
conn_mat_fn = self.params['conn_mat_fn_base'] + '%s.dat' % (conn_type)
delay_mat_fn = self.params['delay_mat_fn_base'] + '%s.dat' % (conn_type)
if os.path.exists(conn_mat_fn):
print 'Loading', conn_mat_fn
self.connection_matrices[conn_type] = np.loadtxt(conn_mat_fn)
# delays_ee = np.loadtxt(delay_mat_ee_fn)
else:
self.connection_matrices[conn_type], self.delays[conn_type] = utils.convert_connlist_to_matrix(params['merged_conn_list_%s' % conn_type], n_src, n_tgt)
np.savetxt(conn_mat_fn, self.connection_matrices[conn_type])
# np.savetxt(delay_mat_fn, self.delays[conn_type])
if conn_type == 'ee':
self.conn_mat_loaded[0] = True
elif conn_type == 'ei':
self.conn_mat_loaded[1] = True
elif conn_type == 'ie':
self.conn_mat_loaded[2] = True
elif conn_type == 'ii':
self.conn_mat_loaded[3] = True
elif conn_type == 'ie':
n_src, n_tgt = params['n_inh'], params['n_exc']
elif conn_type == 'ii':
n_src, n_tgt = params['n_inh'], params['n_inh']
conn_list_fn = params['merged_conn_list_%s' % conn_type]
print 'Loading: ', conn_list_fn
conn_mat_fn = params['conn_mat_fn_base'] + '%s.dat' % (conn_type)
#delay_mat_fn = params['delay_mat_fn_base'] + '%s.dat' % (conn_type)
if os.path.exists(conn_mat_fn):
print 'Loading', conn_mat_fn
w = np.loadtxt(conn_mat_fn)
# delays_ee = np.loadtxt(delay_mat_ee_fn)
else:
w, delays = utils.convert_connlist_to_matrix(params['merged_conn_list_%s' % conn_type], n_src, n_tgt)
print 'Saving:', conn_mat_fn
np.savetxt(conn_mat_fn, w)
print 'Weights: min %.2e median %.2e max %.2e mean %.2e st %.2e' % (w.min(), w.max(), np.median(w), w.mean(), w.std())
fig = pylab.figure()
ax = fig.add_subplot(111)
print "plotting ...."
title = 'Connection matrix %s \nw_sigma_x(v): %.1f (%.1f)' % (conn_type, params['w_sigma_x'], params['w_sigma_v'])
ax.set_title(title)
cax = ax.pcolormesh(w)#, edgecolor='k', linewidths='1')
ax.set_ylim(0, w.shape[0])
ax.set_xlim(0, w.shape[1])
ax.set_ylabel('Target')
ax.set_xlabel('Source')
pylab.colorbar(cax)
def get_cond_matrix(nspikes, w):
cond_matrix = np.zeros((params['n_exc'], params['n_exc']))
for tgt in xrange(params['n_exc']):
for src in xrange(params['n_exc']):
if src != tgt:
cond_matrix[src, tgt] = w[src, tgt] * nspikes[src]
return cond_matrix
# compute everything
input_cond = calculate_input_cond()
os.system("python merge_connlists.py")
conn_list_fn = params['merged_conn_list_ee']
print 'debug', conn_list_fn
w, delays = utils.convert_connlist_to_matrix(conn_list_fn, params['n_exc'], params['n_exc'])
cond_matrix = get_cond_matrix(nspikes, w)
np.savetxt(params['tmp_folder'] + 'input_cond.dat', input_cond)
np.savetxt(params['tmp_folder'] + 'cond_matrix.dat', cond_matrix)
# or load the computed cond_matrix etc (for replotting)
#input_cond = np.loadtxt(params['tmp_folder'] + 'input_cond.dat')
#cond_matrix = np.loadtxt(params['tmp_folder'] + 'cond_matrix.dat')
summed_network_cond = np.zeros(params['n_exc'])
for tgt in xrange(params['n_exc']):
summed_network_cond[tgt] = cond_matrix[:, tgt].sum()
import utils import simulation_parameters import CreateConnections as CC network_params = simulation_parameters.parameter_storage() # network_params class containing the simulation parameters params = network_params.load_params() # params stores cell numbers, etc as a dictionary tp = np.loadtxt(params['tuning_prop_means_fn']) mp = params['motion_params'] print "Motion parameters", mp sigma_x, sigma_v = params['w_sigma_x'], params['w_sigma_v'] # small sigma values let p and w shrink print 'utils.sort_gids_by_distance_to_stimulus...' indices, distances = utils.sort_gids_by_distance_to_stimulus(tp , mp) # cells in indices should have the highest response to the stimulus print 'utils.convert_connlist_to_matrix...' conn_mat, delays = utils.convert_connlist_to_matrix(params['conn_list_ee_fn_base'] + '0.dat', params['n_exc'], params['n_exc']) #n = 50 n = int(params['n_exc'] * .05) # fraction of 'interesting' cells print "Loading nspikes", params['exc_spiketimes_fn_merged'] + '.ras' nspikes = utils.get_nspikes(params['exc_spiketimes_fn_merged'] + '.ras', n_cells=params['n_exc']) mans = (nspikes.argsort()).tolist() # mans = most active neurons mans.reverse() mans = mans[0:n] print 'w_out_good = weights to other well tuned neurons' print 'w_in_good = weights from other well tuned neurons' print 'w_out_total = sum of all outgoing weights' print 'w_in_total = sum of all incoming weights' print 'distance_to_stim = minimal distance to the moving stimulus (linear sum of (time-varying) spatial distance and (constant) distance in velocity)' print "\nGID\tnspikes\tdistance_to_stim\tw_out_good\tw_in_good\tw_out_total\tw_in_total\ttuning_prop"
# cax = ax.pcolor(data)#, edgecolor='k', linewidths='1')
# ax.set_ylim(0, data.shape[0])
# ax.set_xlim(0, data.shape[1])
# pylab.colorbar(cax)
#cax = ax.pcolor(data, cmap='binary')
#cax = ax.pcolor(data, cmap='RdBu')
#cax = ax.imshow(data[:,:12])
if (n_dw > 0):
# dws = [np.zeros((n_cells, n_cells)) for i in xrange(n_dw)]
# plot the difference weight matrix
for i in xrange(len(fns)-1):
fn1 = fns[i]
fn2 = fns[i+1]
d1 = utils.convert_connlist_to_matrix(fn1, n_cells)
d2 = utils.convert_connlist_to_matrix(fn2, n_cells)
dw = d2 - d1
data = dw
print "plotting dw"
fig = pylab.figure()
ax = fig.add_subplot(121)
ax.set_title("Difference %s \n- %s" % (fn1, fn2))
cax = ax.pcolor(data)#, edgecolor='k', linewidths='1')
#cax = ax.pcolor(data, cmap='binary')
#cax = ax.pcolor(data, cmap='RdBu')
#cax = ax.imshow(data[:,:12])
ax.set_ylim(0, data.shape[0])
ax.set_xlim(0, data.shape[1])
pylab.colorbar(cax)