CODEWITHSUNDEEP
tensorflowkeras
user7867665
user7867665

Reputation: 882

Multiply a layer with 1 output to a layer with multiple outputs

I am trying to get the neural network to multiply two input tensors (I cannot do it before feeding them into the network). Keras' multiply function can only handle two tensors of the same dimension. Have something like:

scale_, mean_ = 2., 4.
a = Input(shape=(300,), name='Input_vec')
m_num = Input(shape=(1,), name='Input_num')
mulNum = Lambda(lambda x: K.exp(x * scale_ + mean_))(m_num)
output = multiply([mulNum, a]) # tensors not same shape

how can I do the multiplication of two inputs where one is just a scalar?

Upvotes: 1

Views: 115

Answers (1)

Vlad
Vlad

Reputation: 8585

Use tf.multiply() (or tf.math.multiply()) that supports broadcasting:

output = Lambda(lambda x: tf.multiply(x[0], x[1]))((a, mulNum))

Upvotes: 1

Related Questions