Row wise normalization of tensor
rayjs opened this issue · 1 comments
rayjs commented
Hi,
I am new to mshadow. I am seeking some help regarding a tensor operation. I see there are examples to implement elementwise tensor operations as below:
https://github.com/dmlc/mshadow/tree/3d87ed2a4b47ef749c616f208cee45d920fb6e6e/guide#user-defined-operator
But I am looking for slightly different that is, row wise normalization of a tensor
A= [row1; row2;... rowN]; Anormalized = [row1/norm(row1); ....; rowN/norm(rowN)];
norm refers to the 2-norm.
This can be also thought of as reduction and then broadcast along the rows of the tensor.
Is there any implementation or any pointer would be helpful?