You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Is your feature request related to a problem? Please describe.
The FunFact tensor expression syntax lacks a way to express convolutions and tensor reshaping in a convenient manner.
Describe the solution you'd like
We could use index addition to denote convolution, e.g.
A[i+p, j+q] B[p, q] #convolving A with B using B as a filter.
Following a similar line of thought, index multiplication can denote both reshaping and Kronecker product (since Kronecker product can be thought of as outer product + reshaping):
A[i] B[j] >> [i*j] # Kronecker product between A and B
A[i, j, k] >> [i*j, k] #reshape two first dimensions of A to a vector
Describe alternatives you've considered
Currently, we have the unary * operator to denote Kronecker product, e.g.:
A[[*i]] B[[*i]] # Kronecker product between A
but we do not have the means to express reshaping or convolution yet.
Is your feature request related to a problem? Please describe.
The FunFact tensor expression syntax lacks a way to express convolutions and tensor reshaping in a convenient manner.
Describe the solution you'd like
We could use index addition to denote convolution, e.g.
Following a similar line of thought, index multiplication can denote both reshaping and Kronecker product (since Kronecker product can be thought of as outer product + reshaping):
Describe alternatives you've considered
Currently, we have the unary
*
operator to denote Kronecker product, e.g.:but we do not have the means to express reshaping or convolution yet.
Additional context
#216.
The text was updated successfully, but these errors were encountered: