def test_mult_by_sigmoid_deriv():
m = 256
n = 128
c_targets = np.array(np.random.randn(m, n)*10, dtype=np.float32, order='F')
c_acts = np.array(np.random.rand(m, n), dtype=np.float32, order='F')
g_targets = cm.CUDAMatrix(c_targets)
g_acts = cm.CUDAMatrix(c_acts)
c_targets = c_targets * c_acts * (1. - c_acts)
cl.mult_by_sigmoid_deriv(g_targets, g_acts)
assert np.max(np.abs(c_acts - g_acts.asarray())) < 10**-2, "Error in cudamat.learn.mult_by_sigmoid_deriv exceeded threshold"
def test_mult_by_sigmoid_deriv():
m = 256
n = 128
c_targets = np.array(np.random.randn(m, n) * 10,
dtype=np.float32,
order='F')
c_acts = np.array(np.random.rand(m, n), dtype=np.float32, order='F')
g_targets = cm.CUDAMatrix(c_targets)
g_acts = cm.CUDAMatrix(c_acts)
c_targets = c_targets * c_acts * (1. - c_acts)
cl.mult_by_sigmoid_deriv(g_targets, g_acts)
assert np.max(
np.abs(c_acts - g_acts.asarray())
) < 10**-2, "Error in cudamat.learn.mult_by_sigmoid_deriv exceeded threshold"
def backProp(self, error):
# print 'back propagation'
self.dW[self.H-1].add_dot(self.vis[self.H-1],error.T)
# print 'self.vis'
# self.vis[self.H-1].copy_to_host()
# print self.vis[self.H-1].numpy_array
# print 'self.dW'
# self.dW[self.H-1].copy_to_host()
# print self.dW[self.H-1].numpy_array
# print 'error 2'
# error.copy_to_host()
# print error.numpy_array
self.db[self.H-1].add_sums(error,axis =1 )
for i in list(reversed(range(self.H-1))):
delta = cm.empty((self.W[i+1].shape[0],error.shape[1]))
cm.dot(self.W[i+1],error,target = delta)# delta : 2000*256
learn.mult_by_sigmoid_deriv(delta, self.vis[i+1])
self.dW[i].add_dot(self.vis[i], delta.T)
self.db[i].add_sums(delta, axis = 1)
error = delta
