You need to use theano's dimshuffle to add a the "1" dimension to the vector then it will broadcast correctly.
out = v.dimshuffle(0, 'x') * B where the "0" in dimshuffle represents the first (and only) axis of the vector and "'x'" represents a new broadcastable dimension of length 1. On Tuesday, August 30, 2016 at 10:25:26 PM UTC-7, Haohan Wang wrote: > > Hi, > > Is there any way in Theano that supports a vector broadcasting to a > matrix? The document does not indicate so, but I can hardly believe a tool > like theano does not support this. > > Basically, what I need is: > > a vector *v *of size n x 1, and a matrix *B *of size (n x m), and I hope > to take the product of *v* and *B* that returns the result as a (*n** x m*) > matrix, where each column *i *of the result is an element-wise product of > *v *and *B_i (1 < i < m)* > > Thanks. > Haohan > -- --- You received this message because you are subscribed to the Google Groups "theano-users" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/d/optout.
