https://github.com/jonascarpay/convoluted

Dependently typed convolutional neural networks
https://github.com/jonascarpay/convoluted

convolutional-neural-networks dependent-types haskell

Last synced: 4 months ago
JSON representation

Dependently typed convolutional neural networks

• Host: GitHub
• URL: https://github.com/jonascarpay/convoluted
• Owner: jonascarpay
• License: bsd-3-clause
• Created: 2017-02-07T22:19:05.000Z (over 7 years ago)
• Default Branch: master
• Last Pushed: 2017-05-10T09:38:38.000Z (about 7 years ago)
• Last Synced: 2023-10-20T20:49:04.912Z (8 months ago)
• Topics: convolutional-neural-networks, dependent-types, haskell
• Language: Haskell
• Homepage:
• Size: 98.6 KB
• Stars: 33
• Watchers: 4
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Lists

• awesome-haskell-deep-learning - convoluted - Dependently typed convolutional neural networks in pure Haskell. Uses the repa library for high-performance arrays, with a static wrapper that ensures networks are valid at compile-time. | [Jonas Carpay](https://github.com/jonascarpay) (Haskell Packages / Legacy Packages)

README

        
# convoluted

Dependently typed convolutional neural networks in pure Haskell.

Uses the [repa library](https://hackage.haskell.org/package/repa) for high-performance arrays,

with a static wrapper that ensures networks are valid at compile-time.

### Example

```haskell

type BatchSize = 100

type MNIST = Network (ZZ ::. BatchSize ::. 1 ::. 28 ::. 28)

                     '[ Convolution 5 1 13 13 16 16

                      , Pool

                      , ReLU

                      , Flatten

                      , FC 320 10

                      , MultiSoftMax '[10] ]

                     (ZZ ::. BatchSize ::. 10)

mnist = randomNetwork seed :: MNIST

```

see [visor](https://github.com/jonascarpay/visor) or [convoluted-mnist](https://github.com/jonascarpay/convoluted-mnist) for examples.

#### Thanks

I borrowed heavily from [Huw Campbell's grenade](https://github.com/HuwCampbell/grenade), and like the author of that library, from [Justin Le's blog posts on dependently typed neural networks](https://blog.jle.im/entry/practical-dependent-types-in-haskell-1.html).