def update_c_est(self, scale=1, regress=False):
if regress:
#For regression, shape M to (TxD), for each episode (so S x (TxD))
#(reshape by going through each muscle and then to next)
M_flat = np.reshape(self.M_stacked, (self.S, self.T * self.D))
#Each column of predictor is one synergy, shape (TxD)
#predictor has N columns
W_flat = np.reshape(self.W_est_stacked,
(self.N, self.T * self.D)).T
#loop through and do a regression for each episode
for s in range(self.S):
self.c_est[s, :], _ = scipy.optimize.nnls(W_flat, M_flat[s, :])
else:
mult_factor = np.zeros_like(self.c_est)
for s in range(self.S):
for i in range(self.N):
Theta_i_tis = self.Theta[
i,
util.t_shift_to_index(self.delays[s, i], self.T), :, :]
num = np.trace((self.M_stacked[s, :, :].T).dot(
self.W_est_spread).dot(Theta_i_tis))
denom = np.trace((self.H_stacked[s, :, :].T).dot(
self.W_est_spread.T).dot(
self.W_est_spread).dot(Theta_i_tis))
mult_factor[s, i] = num / denom
self.c_est = self.c_est * scale * mult_factor
def update_H(self):
H = np.zeros((self.S, self.N * self.T, 2 * self.T))
for s in range(self.S):
H[s,:,:] = np.sum(self.c_est[s,:][:,None,None]*\
np.array([self.Theta[i,util.t_shift_to_index(
self.delays[s,i], self.T),:,:] for i in range(self.N)]),axis=0)
self.H_stacked = H
self.H_spread = np.concatenate(H, axis=1)
def construct_Theta(N, T):
Theta = np.zeros(
(N, 2 * T - 1, N * T, T)) #shape of each Theta_i(t) is N*T x T
for i in range(1, N + 1):
for t in range(1 - T, T):
rows, columns = np.indices((N * T, T))
to_fill = (rows + 1 - (i - 1) * T) == (columns + 1 - t)
to_fill[0:(i - 1) * T, :] = 0.
to_fill[i * T:, :] = 0.
Theta[i - 1, util.t_shift_to_index(t, T), :, :] = to_fill
return Theta
def multiplicative_update_c(c_est, M, W_est, Theta, H, delays, scale=1):
_, _, _, T = np.shape(Theta)
S, N = np.shape(c_est)
M = util.stack(M, T)
# plt.figure(9)
# plt.imshow(H)
# print(np.shape(H),(N*T,T*S))
H = util.stack(H, T)
# print(np.shape(H),(S,N*T,T))
# plt.figure(10)
# plt.imshow(H[0,:,:])
# plt.show()
mult_factor = np.zeros_like(c_est)
for s in range(S):
for i in range(N):
# print(delays[s,i])
Theta_i_tis = Theta[i,
util.t_shift_to_index(delays[s, i], T), :, :]
# print(np.sum(Theta_i_tis))
# raw_input(' ')
num = np.trace((M[s, :, :].T).dot(W_est).dot(Theta_i_tis))
denom = np.trace(
(H[s, :, :].T).dot(W_est.T).dot(W_est).dot(Theta_i_tis))
# if denom==0:
# print(s,i)
# print('inf denom c update')
# plt.figure(55)
# plt.subplot(2,1,1)
# plt.imshow((H[s,:,:].T).dot(W_est.T).dot(W_est).dot(Theta_i_tis),
# interpolation='none')
# plt.subplot(2,1,2)
# plt.imshow((H[s,:,:]),interpolation='none')
# plt.show()
# raw_input(' ')
mult_factor[s, i] = num / denom
# print(mult_factor)
c_est = c_est * scale * mult_factor
return c_est
cmap='Greys_r',
vmin=0,
vmax=amp_max)
pltuls.strip_ticks(ax)
plt.text(0.65, 0.95, 'True M', transform=plt.gcf().transFigure)
Theta = np.zeros(
(N, 2 * T - 1, N * T, T)) #shape of each Theta_i(t) is N*T x T
for i in range(1, N + 1):
for t in range(1 - T, T):
rows, columns = np.indices((N * T, T))
to_fill = (rows + 1 - (i - 1) * T) == (columns + 1 - t)
to_fill[0:(i - 1) * T, :] = 0.
to_fill[i * T:, :] = 0.
Theta[i - 1, util.t_shift_to_index(t, T), :, :] = to_fill
# plt.figure(544)
# plt.imshow(Theta[i-1,util.t_shift_to_index(t,T),:,:],interpolation='none')
# raw_input(' ')
# util.test_Theta(Theta)
# error = np.inf
unexp_var_threshold = 1e-5
# R2_diff_threshold = 1e-7
# R2 = np.zeros(2)
SS_tot = substeps.compute_total_sum_squares(M)
print(SS_tot)
SS_res = substeps.compute_squared_error(
W_est, c_est, np.zeros_like(c_est),