Method __init__

rl2/cartpole/pg_tf.py:120–147

(self, D, hidden_layer_sizes)

# approximates V(s)
class ValueModel:
  def __init__(self, D, hidden_layer_sizes):
    # create the graph
    self.layers = []
    M1 = D
    for M2 in hidden_layer_sizes:
      layer = HiddenLayer(M1, M2)
      self.layers.append(layer)
      M1 = M2

    # final layer
    layer = HiddenLayer(M1, 1, lambda x: x)
    self.layers.append(layer)

    # inputs and targets
    self.X = tf.placeholder(tf.float32, shape=(None, D), name='X')
    self.Y = tf.placeholder(tf.float32, shape=(None,), name='Y')

    # calculate output and cost
    Z = self.X
    for layer in self.layers:
      Z = layer.forward(Z)
    Y_hat = tf.reshape(Z, [-1]) # the output
    self.predict_op = Y_hat

    cost = tf.reduce_sum(tf.square(self.Y - Y_hat))
    # self.train_op = tf.train.AdamOptimizer(1e-2).minimize(cost)
    # self.train_op = tf.train.MomentumOptimizer(1e-2, momentum=0.9).minimize(cost)
    self.train_op = tf.train.GradientDescentOptimizer(1e-4).minimize(cost)

  def set_session(self, session):
    self.session = session