https://github.com/mstksg/hmatrix-backprop

backprop primitives for hmatrix
https://github.com/mstksg/hmatrix-backprop

Last synced: about 2 months ago
JSON representation

backprop primitives for hmatrix

• Host: GitHub
• URL: https://github.com/mstksg/hmatrix-backprop
• Owner: mstksg
• License: bsd-3-clause
• Created: 2018-02-05T21:49:46.000Z (almost 7 years ago)
• Default Branch: master
• Last Pushed: 2020-02-04T19:00:07.000Z (almost 5 years ago)
• Last Synced: 2024-10-13T15:43:22.279Z (2 months ago)
• Language: Haskell
• Homepage: http://hackage.haskell.org/package/hmatrix-backprop
• Size: 798 KB
• Stars: 13
• Watchers: 6
• Forks: 3
• Open Issues: 2
• Metadata Files:
  • Readme: README.md
  • Changelog: CHANGELOG.md
  • License: LICENSE

Awesome Lists containing this project

README

        
hmatrix-backprop

================

[![hmatrix-backprop on Hackage](https://img.shields.io/hackage/v/hmatrix-backprop.svg?maxAge=2592000)](https://hackage.haskell.org/package/hmatrix-backprop)

[![Build Status](https://travis-ci.org/mstksg/hmatrix-backprop.svg?branch=master)](https://travis-ci.org/mstksg/hmatrix-backprop)

*[hmatrix][]* operations lifted for *[backprop][]*, along with orphan instances

of `Backprop`.

[hmatrix]: http://hackage.haskell.org/package/hmatrix

[backprop]: http://hackage.haskell.org/package/backprop

Meant to act as a drop-in replacement to the API of

[Numeric.LinearAlgebra.Static][static].  Just change your imports, and your

functions are automatically backpropagatable.  Useful types are all

re-exported.

[static]: https://hackage.haskell.org/package/hmatrix-0.18.2.0/docs/Numeric-LinearAlgebra-Static.html

Formulas for gradients come from the following papers:

*   

*   

*   

*   

Some functions are not yet implemented!  See module documentation for details.

PR's definitely appreciated :)

Tests

-----

Currently numeric tests are implemented as property tests using hedgehog, but

it is possible that the answers might differ from the true values by an amount

undetectable by property tests.

All functions currently are tested except for the higher-order functions.

They are tested by "nudging" components of inputs and checking if the change in

the function outputs match what is expected from the backpropagated gradient.

TODO

----

1.  Now that new backprop no longer requires `Num`, we can lift normal hmatrix

    operations as well.

1.  Statically sized convolutions.  Should probably add this to hmatrix instead

    first, though.