T-net code problem

[milk-bottle-liyu]

Hi, Im quite new to tensorflow. I read your code and notice here: ` with tf.variable_scope('transform_feat') as sc:

    weights = tf.get_variable('weights', [256, K*K],
                              initializer=tf.constant_initializer(0.0),
                              dtype=tf.float32)

    biases = tf.get_variable('biases', [K*K],
                             initializer=tf.constant_initializer(0.0),
                             dtype=tf.float32)

    biases += tf.constant(np.eye(K).flatten(), dtype=tf.float32)

    transform = tf.matmul(net, weights)

    transform = tf.nn.bias_add(transform, biases)

` if u initialize "weights" with 0, should not it stay 0 for ever?

[LoFaiTh]

I have the same question. Have any one figure this out? Can you explain this to us? @charlesq34

[ameliajimenez]

Hi, no, it won't.

That's just the initialization of the weights of that layer. They will be updated after computing the gradients of each batch, like any other weights from convolutional layers.