On Thu, Nov 03, 2016, Nick wrote: > I think I can do x[x < T.max(x,axis=-1)] = 0 .. but theano shared doesnt > support assignment
Theano does not support assignment, you have to use T.set_subtensor instead, and used the returned value. Moreover, indexing with a mask is not supported yet, but you can use nonzeros(). For instance, your case would look something like: y = T.set_subtensor(x[(x < x.max(axis=-1, keepdims=True)).nonzeros()], 0) > > On Thursday, November 3, 2016 at 8:49:25 PM UTC-4, Nick wrote: > > > > For example, > > > > x = [[1,2],[3,4]] => [[0,0],[3,4]] > > > > I'm trying to do this but cant seem to get broadcastables right.. > > > > I know columnwise max would be T.max(x,axis=-1) .. What to do from there? > > > > -- > > --- > 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. -- Pascal -- --- 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.
