https://github.com/johannesmessner/lambdagrad

Automatic gradients in ~70 lines of Haskell
https://github.com/johannesmessner/lambdagrad

Last synced: 6 days ago
JSON representation

Automatic gradients in ~70 lines of Haskell

• Host: GitHub
• URL: https://github.com/johannesmessner/lambdagrad
• Owner: JohannesMessner
• Created: 2023-05-22T20:09:34.000Z (over 1 year ago)
• Default Branch: master
• Last Pushed: 2023-05-22T20:20:26.000Z (over 1 year ago)
• Last Synced: 2025-01-07T22:06:11.129Z (13 days ago)
• Language: Haskell
• Size: 1.95 KB
• Stars: 0
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# lambdagrad

Automatic gradients in ~70 lines of Haskell

Load the `Engine` module:

```terminal

$ ghci

GHCi, version 8.10.7: https://www.haskell.org/ghc/  :? for help

Prelude> :l Engine

[1 of 2] Compiling Data             ( Data.hs, interpreted )

[2 of 2] Compiling Engine           ( Engine.hs, interpreted )

Ok, two modules loaded.

```

Use `val` to creat computational graphs:

```bash

*Engine> expression = ((val 0.5) + (val 3)) ** 2

*Engine> expression

InnerNode Pow 12.25 0.0 (InnerNode Plus 3.5 0.0 (Leaf 0.5 0.0) (Leaf 3.0 0.0)) (Leaf 2.0 0.0)

```

Use `backward` to compute gradients:

```bash

*Engine> backward expression

InnerNode Pow 12.25 1.0 (InnerNode Plus 3.5 7.0 (Leaf 0.5 7.0) (Leaf 3.0 7.0)) (Leaf 2.0 0.0)

```

Use `getLeafs` to get leaf node of the computational graph:

```bash

*Engine> getLeafs $ backward expression

[Leaf 0.5 7.0,Leaf 3.0 7.0,Leaf 2.0 0.0]

```

And `gradient` to extract their gradients:

```bash

*Engine> map gradient (getLeafs $ backward expression)

[7.0,7.0,0.0]

```