How exactly does LSTMCell from TensorFlow operates?
i examined this link and your code is almost perfect but you forgot to add forget_bias value(default 1.0) in this line F = vsigmoid(g3) its actualy F = vsigmoid(g3+self.forget_bias) or in your case its 1 F = vsigmoid(g3+1)
here is my imp with numpy:
import numpy as np
import tensorflow as tf
num_units = 3
lstm = tf.nn.rnn_cell.LSTMCell(num_units = num_units)
batch=1
timesteps = 7
num_input = 4
X = tf.placeholder("float", [batch, timesteps, num_input])
x = tf.unstack(X, timesteps, 1)
outputs, states = tf.contrib.rnn.static_rnn(lstm, x, dtype=tf.float32)
sess = tf.Session()
init = tf.global_variables_initializer()
sess.run(init)
x_val = np.reshape(range(28),[batch, timesteps, num_input])
res = sess.run(outputs, feed_dict = {X:x_val})
for e in res:
print(e)
print("\nmy imp\n")
#my impl
def sigmoid(x):
return 1/(1+np.exp(-x))
kernel,bias=sess.run([lstm._kernel,lstm._bias])
f_b_=lstm._forget_bias
c,h=np.zeros([batch,num_input-1]),np.zeros([batch,num_input-1])
for step in range(timesteps):
inpt=np.split(x_val,7,1)[step][0]
lstm_mtrx=np.matmul(np.concatenate([inpt,h],1),kernel)+bias
i,j,f,o=np.split(lstm_mtrx,4,1)
c=sigmoid(f+f_b_)*c+sigmoid(i)*np.tanh(j)
h=sigmoid(o)*np.tanh(c)
print(h)
output:
[[ 0.06964055 -0.06541953 -0.00682676]]
[[ 0.005264 -0.03234607 0.00014838]]
[[ 1.617855e-04 -1.316892e-02 8.596722e-06]]
[[ 3.9425286e-06 -5.1347450e-03 7.5078127e-08]]
[[ 8.7508155e-08 -1.9560163e-03 6.3853928e-10]]
[[ 1.8867894e-09 -7.3784427e-04 5.8551406e-12]]
[[ 4.0385355e-11 -2.7728223e-04 5.3957669e-14]]
my imp
[[ 0.06964057 -0.06541953 -0.00682676]]
[[ 0.005264 -0.03234607 0.00014838]]
[[ 1.61785520e-04 -1.31689185e-02 8.59672610e-06]]
[[ 3.94252745e-06 -5.13474567e-03 7.50781122e-08]]
[[ 8.75080644e-08 -1.95601574e-03 6.38539112e-10]]
[[ 1.88678843e-09 -7.37844070e-04 5.85513438e-12]]
[[ 4.03853841e-11 -2.77282006e-04 5.39576024e-14]]
Tensorflow uses glorot_uniform() function to initialize the lstm kernel, which samples weights from a random uniform distribution. We need to fix a value for the kernel to get reproducible results:
import tensorflow as tf
import numpy as np
np.random.seed(0)
timesteps = 7
num_input = 4
x_val = np.random.normal(size = (1, timesteps, num_input))
num_units = 3
def glorot_uniform(shape):
limit = np.sqrt(6.0 / (shape[0] + shape[1]))
return np.random.uniform(low=-limit, high=limit, size=shape)
kernel_init = glorot_uniform((num_input + num_units, 4 * num_units))
My implementation of the LSTMCell (well, actually it's just slightly rewritten tensorflow's code):
def sigmoid(x):
return 1. / (1 + np.exp(-x))
class LSTMCell():
"""Long short-term memory unit (LSTM) recurrent network cell.
"""
def __init__(self, num_units, initializer=glorot_uniform,
forget_bias=1.0, activation=np.tanh):
"""Initialize the parameters for an LSTM cell.
Args:
num_units: int, The number of units in the LSTM cell.
initializer: The initializer to use for the kernel matrix. Default: glorot_uniform
forget_bias: Biases of the forget gate are initialized by default to 1
in order to reduce the scale of forgetting at the beginning of
the training.
activation: Activation function of the inner states. Default: np.tanh.
"""
# Inputs must be 2-dimensional.
self._num_units = num_units
self._forget_bias = forget_bias
self._activation = activation
self._initializer = initializer
def build(self, inputs_shape):
input_depth = inputs_shape[-1]
h_depth = self._num_units
self._kernel = self._initializer(shape=(input_depth + h_depth, 4 * self._num_units))
self._bias = np.zeros(shape=(4 * self._num_units))
def call(self, inputs, state):
"""Run one step of LSTM.
Args:
inputs: input numpy array, must be 2-D, `[batch, input_size]`.
state: a tuple of numpy arrays, both `2-D`, with column sizes `c_state` and
`m_state`.
Returns:
A tuple containing:
- A `2-D, [batch, output_dim]`, numpy array representing the output of the
LSTM after reading `inputs` when previous state was `state`.
Here output_dim is equal to num_units.
- Numpy array(s) representing the new state of LSTM after reading `inputs` when
the previous state was `state`. Same type and shape(s) as `state`.
"""
num_proj = self._num_units
(c_prev, m_prev) = state
input_size = inputs.shape[-1]
# i = input_gate, j = new_input, f = forget_gate, o = output_gate
lstm_matrix = np.hstack([inputs, m_prev]).dot(self._kernel)
lstm_matrix += self._bias
i, j, f, o = np.split(lstm_matrix, indices_or_sections=4, axis=0)
# Diagonal connections
c = (sigmoid(f + self._forget_bias) * c_prev + sigmoid(i) *
self._activation(j))
m = sigmoid(o) * self._activation(c)
new_state = (c, m)
return m, new_state
X = x_val.reshape(x_val.shape[1:])
cell = LSTMCell(num_units, initializer=lambda shape: kernel_init)
cell.build(X.shape)
state = (np.zeros(num_units), np.zeros(num_units))
for i in range(timesteps):
x = X[i,:]
output, state = cell.call(x, state)
print(output)
Produces output:
[-0.21386017 -0.08401277 -0.25431477]
[-0.22243588 -0.25817422 -0.1612211 ]
[-0.2282134 -0.14207162 -0.35017249]
[-0.23286737 -0.17129192 -0.2706512 ]
[-0.11768674 -0.20717363 -0.13339118]
[-0.0599215 -0.17756104 -0.2028935 ]
[ 0.11437953 -0.19484555 0.05371994]
While your Tensorflow code, if you replace the second line with
lstm = tf.nn.rnn_cell.LSTMCell(num_units = num_units, initializer = tf.constant_initializer(kernel_init))
returns:
[[-0.2138602 -0.08401276 -0.25431478]]
[[-0.22243595 -0.25817424 -0.16122109]]
[[-0.22821338 -0.1420716 -0.35017252]]
[[-0.23286738 -0.1712919 -0.27065122]]
[[-0.1176867 -0.2071736 -0.13339119]]
[[-0.05992149 -0.177561 -0.2028935 ]]
[[ 0.11437953 -0.19484554 0.05371996]]
Considering Linear Algebra, it's possible to exist a dimension mismatch in the matrix multiplication between I*N (red circle), affecting the output, given that n x m dot m x p will give you a n x p dimensional output.
Here is a blog which will answer any conceptual questions related to LSTM's. Seems that there is a lot which goes into building an LSTM from scratch!
Of course, this answer doesn't solve your question but just giving a direction.