def __init__(self, num_visibles, num_hiddens): """ Initializes the parameters of the SemiRBM. @type num_visibles: integer @param num_visibles: number of visible units @type num_hiddens: integer @param num_hiddens: number of hidden units """ AbstractBM.__init__(self, num_visibles, num_hiddens) # additional hyperparameters self.learning_rate_lateral = 0.01 self.momentum_lateral = 0.5 self.weight_decay_lateral = 0. self.damping = 0.2 self.num_lateral_updates = 20 self.sampling_method = AbstractBM.MF # additional parameters self.L = np.matrix(np.random.randn(num_visibles, num_visibles)) / num_visibles / 200 self.L = np.triu(self.L) + np.triu(self.L).T - 2. * np.diag(np.diag(self.L)) self.dL = np.zeros_like(self.L)
def __init__(self, num_visibles, num_hiddens):
AbstractBM.__init__(self, num_visibles, num_hiddens)
# hyperparameters
self.sigma = 0.2
self.vsigma = np.asmatrix(np.ones(num_visibles)).T
self.dvsigma = np.ones_like(self.vsigma)
def __init__(self, num_visibles, num_hiddens): AbstractBM.__init__(self, num_visibles, num_hiddens) # hyperparameters self.sigma = 0.2 self.vsigma = np.asmatrix(np.ones(num_visibles)).T self.dvsigma = np.ones_like(self.vsigma)
def __init__(self, num_visibles, num_hiddens):
"""
Initializes the parameters of the SemiRBM.
@type num_visibles: integer
@param num_visibles: number of visible units
@type num_hiddens: integer
@param num_hiddens: number of hidden units
"""
AbstractBM.__init__(self, num_visibles, num_hiddens)
# additional hyperparameters
self.learning_rate_lateral = 0.01
self.momentum_lateral = 0.5
self.weight_decay_lateral = 0.
self.damping = 0.2
self.num_lateral_updates = 20
self.sampling_method = AbstractBM.MF
# additional parameters
self.L = np.matrix(np.random.randn(num_visibles,
num_visibles)) / num_visibles / 200
self.L = np.triu(self.L) + np.triu(
self.L).T - 2. * np.diag(np.diag(self.L))
self.dL = np.zeros_like(self.L)
def __init__(self, num_visibles, num_hiddens): """ Initializes the parameters of the RBM. @type num_visibles: integer @param num_visibles: number of visible units @type num_hiddens: integer @param num_hiddens: number of hidden units """ AbstractBM.__init__(self, num_visibles, num_hiddens)
def train(self, X): X = np.asmatrix(X) if self.sampling_method is AbstractBM.MF: # positive phase self.forward(X) # store posterior probabilities Q = self.Q.copy() if self.persistent: self.X = self.pX self.Y = self.pY # negative phase for t in range(self.cd_steps): self.backward() self.forward() if self.persistent: self.pX = self.X.copy() self.pY = self.Y.copy() # update parameters self.dW = X * Q.T / X.shape[1] - self.P * self.Y.T / self.X.shape[1] \ - self.weight_decay * self.W \ + self.momentum * self.dW self.db = X.mean(1) - self.P.mean(1) + self.momentum * self.db self.dc = Q.mean(1) - self.Y.mean(1) + self.momentum * self.dc self.dL = X * X.T / X.shape[1] - self.P * self.P.T / self.X.shape[1] \ - self.weight_decay_lateral * self.L \ + self.momentum_lateral * self.dL self.dL -= np.diag(np.diag(self.dL)) self.W += self.dW * self.learning_rate self.b += self.db * self.learning_rate self.c += self.dc * self.learning_rate self.L += self.dL * self.learning_rate_lateral else: # update RBM parameters AbstractBM.train(self, X) # update lateral connections self.dL = X * X.T / X.shape[1] - self.X * self.X.T / self.X.shape[1] \ - self.weight_decay_lateral * self.L \ + self.momentum_lateral * self.dL self.dL -= np.diag(np.diag(self.dL)) self.L += self.dL * self.learning_rate_lateral
def train(self, X):
X = np.asmatrix(X)
if self.sampling_method is AbstractBM.MF:
# positive phase
self.forward(X)
# store posterior probabilities
Q = self.Q.copy()
if self.persistent:
self.X = self.pX
self.Y = self.pY
# negative phase
for t in range(self.cd_steps):
self.backward()
self.forward()
if self.persistent:
self.pX = self.X.copy()
self.pY = self.Y.copy()
# update parameters
self.dW = X * Q.T / X.shape[1] - self.P * self.Y.T / self.X.shape[1] \
- self.weight_decay * self.W \
+ self.momentum * self.dW
self.db = X.mean(1) - self.P.mean(1) + self.momentum * self.db
self.dc = Q.mean(1) - self.Y.mean(1) + self.momentum * self.dc
self.dL = X * X.T / X.shape[1] - self.P * self.P.T / self.X.shape[1] \
- self.weight_decay_lateral * self.L \
+ self.momentum_lateral * self.dL
self.dL -= np.diag(np.diag(self.dL))
self.W += self.dW * self.learning_rate
self.b += self.db * self.learning_rate
self.c += self.dc * self.learning_rate
self.L += self.dL * self.learning_rate_lateral
else:
# update RBM parameters
AbstractBM.train(self, X)
# update lateral connections
self.dL = X * X.T / X.shape[1] - self.X * self.X.T / self.X.shape[1] \
- self.weight_decay_lateral * self.L \
+ self.momentum_lateral * self.dL
self.dL -= np.diag(np.diag(self.dL))
self.L += self.dL * self.learning_rate_lateral
def __init__(self, num_visibles, num_hiddens):
AbstractBM.__init__(self, num_visibles, num_hiddens)
# hyperparameters
self.sigma = 1.
def __init__(self, num_visibles, num_hiddens): AbstractBM.__init__(self, num_visibles, num_hiddens) # hyperparameters self.sigma = 1.
