CODEWITHSUNDEEP
keras
Mahedi Hasan
Mahedi Hasan

Reputation: 23

How to set bias value of a particular layer in Keras

I was trying to set bias value of a particular layer in Keras, but I didn't find any way to do that.

For example weight can be set by following code:

model.layers[-1].set_weights(weights)

Is there any way to set bias, the way weights can be set (like above)?

Can anyone help me in this regard?

Upvotes: 2

Views: 4840

Answers (3)

Ashish Jha
Ashish Jha

Reputation: 1

In tensorflow .get_weights() method on layer returns weights and bias of that layer. To set any layer weight and bias just use .set_weights() method. for example to set layer_b weights and bias from layer_a do as follow:

layer_b.set_weights(layer_a.get_weights())

for reference you can refer set_weights

Upvotes: 0

Niousha
Niousha

Reputation: 1

you can do this by setting constraints for biases or weights or both.

in each layer there are bias_constraint and kernel_constraint that they are for biases and weights respectively. Here is the example if you want to set bias value to 1.

model.add(Dense(1 ,bias_constraint=tf.keras.constraints.MinMaxNorm(
min_value=1.0, max_value=1.0, rate=1.0, axis=0), activation='softmax'))

Upvotes: 0

mpariente
mpariente

Reputation: 660

You can use K.set_value for this :
K.set_value(model.layers[-1].weights[1], np.ones((bias_dim,)))

Note that when you use set_weights, you have to provide a list containing both weights and biais, so if you would only want to set the weights without setting the biais, you could also use K.set_value.

Upvotes: 4

Related Questions