States in Embedding Layer

[stuartn60]

Hi ... I'm trying to implement Jason Brownlee's Python example How to Use Word Embedding Layers for Deep Learning with Keras which uses the following Embedding layer with an embedding_matrix of GloVe weights:

model = Sequential() e = Embedding(vocab_size, 100, weights=[embedding_matrix], input_length=4, trainable=False) model.add(e)

Could you please help me understand adding weights to a KerasR layer? In https://github.com/statsmaths/kerasR/issues/5 you suggest applying add_weight after the embedding layer. How should the arguments be specified to apply the GloVe weights in this kerasR code?

mod <- Sequential() mod$add(Embedding(vocab_size,100,input_length=4,input_shape=c(4))) mod$add_weight(???, trainable=FALSE)

For example, name and initializer.

Thanks!

Tensorflow Keras documentation provides these arguments:

• name: String, the name for the weight variable.
• dtype: The dtype of the weight.
• initializer: An Initializer instance (callable).
• regularizer: An optional Regularizer instance.
• trainable: A boolean, whether the weight should be trained via backprop or not (assuming that the layer itself is also trainable).
• constraint: An optional Constraint instance.

[stuartn60]

I cannot see how to use add_weight as you suggest ... However from SiWorgan's comment here there seems to be a workround: mod$layers[[1]]$kernel_initializer=embedding_matrix mod$layers[[1]]$trainable=FALSE I could not get weights to work as in the following: mod$layers[[1]]$weights=embedding_matrix Your thoughts on using add_weight and the above workarounds would be much apreciated.