[GitHub] [incubator-tvm] yzhliu commented on issue #2498: [TVM] Automatic differentiation for tensor expressions

[GitBox <gi...@apache.org>]

yzhliu commented on issue #2498: [TVM] Automatic differentiation for tensor expressions
URL: https://github.com/apache/incubator-tvm/pull/2498#issuecomment-595650436
 
 
   @sgrechanik-h I have difficulty understand [this part](https://github.com/sgrechanik-h/tvm/blob/autodiff-dev/src/pass/zero_elimination.cc#L1641-L1645), would be helpful if you can provide some guide,
   
   say we have `A x = b =>  U^T S V^T x = b` the solution for `x` is `V S^{-1} U b + K z`, in which `K` is the right `m-r` columns of `V` and `z` is arbitrary vector.  in my understanding,
   
   * `old_to_new` is `V^T`
   * `matrix` is `S`
   * `rhs` is `U b`
   * `solution[]` in the code is  `V^T V S^{-1} U b + V^T K z`
   
   Then the for-loops in the code calculates `V^T * solution`, how does it equal to original `x` ?
   

----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
users@infra.apache.org


With regards,
Apache Git Services