Weight multiplication layer

[sayan ghosh]

Hi,

I envision a layer which is similar to fully connected layer, but where given N input neurons x_1, x_2,....x_N, there will be N outputs y_1,y_2,....y_N, such that, y_i=w_i*x_i, where {w_i} will consist of the learned weights. Is there an existing layer supporting this, or one which can be modified very slightly to produce this ? This might arise in unbalanced problems, where each class weight needs to be adjusted separately.

Regards,

Sayan