hmatrix-backprop: hmatrix operations lifted for backprop
hmatrix operations lifted for backprop, along with orphan instances.
Meant to act as a drop-in replacement to the API of Numeric.LinearAlgebra.Static. Just change your imports, and your functions are automatically backpropagatable.
See README on Github at https://github.com/mstksg/hmatrix-backprop#readme
[Skip to Readme]
Downloads
- hmatrix-backprop-0.1.3.0.tar.gz [browse] (Cabal source package)
- Package description (as included in the package)
Maintainer's Corner
For package maintainers and hackage trustees
Candidates
- No Candidates
| Versions [RSS] | 0.1.0.0, 0.1.1.0, 0.1.2.0, 0.1.2.1, 0.1.2.2, 0.1.2.3, 0.1.2.4, 0.1.2.5, 0.1.3.0 |
|---|---|
| Change log | CHANGELOG.md |
| Dependencies | backprop (>=0.2.5), base (>=4.7 && <5), ghc-typelits-knownnat, ghc-typelits-natnormalise, hmatrix (>=0.18), hmatrix-vector-sized (>=0.1.1), microlens, vector, vector-sized (>=0.6), vinyl (>=0.9) [details] |
| Tested with | ghc >=8.0 |
| License | BSD-3-Clause |
| Copyright | (c) Justin Le 2018 |
| Author | Justin Le |
| Maintainer | justin@jle.im |
| Category | Math |
| Home page | https://github.com/mstksg/hmatrix-backprop#readme |
| Bug tracker | https://github.com/mstksg/hmatrix-backprop/issues |
| Source repo | head: git clone https://github.com/mstksg/hmatrix-backprop |
| Uploaded | by jle at 2019-08-18T07:28:11Z |
| Distributions | LTSHaskell:0.1.3.0, NixOS:0.1.3.0, Stackage:0.1.3.0 |
| Reverse Dependencies | 1 direct, 0 indirect [details] |
| Downloads | 5031 total (23 in the last 30 days) |
| Rating | (no votes yet) [estimated by Bayesian average] |
| Your Rating | |
| Status | Docs uploaded by user [build log] Last success reported on 2019-08-18 [all 1 reports] |
Readme for hmatrix-backprop-0.1.3.0
[back to package description]hmatrix-backprop
hmatrix operations lifted for backprop, along with orphan instances
of Backprop.
Meant to act as a drop-in replacement to the API of Numeric.LinearAlgebra.Static. Just change your imports, and your functions are automatically backpropagatable. Useful types are all re-exported.
Formulas for gradients come from the following papers:
- https://people.maths.ox.ac.uk/gilesm/files/NA-08-01.pdf
- http://www.dtic.mil/dtic/tr/fulltext/u2/624426.pdf
- http://www.cs.cmu.edu/~zkolter/course/15-884/linalg-review.pdf
- https://arxiv.org/abs/1602.07527
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
- Now that new backprop no longer requires
Num, we can lift normal hmatrix operations as well. - Statically sized convolutions. Should probably add this to hmatrix instead first, though.