/haskell-ml

Machine Learning in Haskell

Primary LanguageHaskellApache License 2.0Apache-2.0

haskell-ml

Machine Learning in Haskell with a focus on neural networks.

NOTE: This was mostly an experiment to see how suitable Haskell would be for machine learning related tasks. I am not planning on continuing work on this as there are very mature deep learning frameworks available nowadays.

Features

  • High speed due to hmatrix

Models

Neutral Networks

  • Multilayer perceptrons

Layers

  • linear
  • hyperbolic
  • sigmoid

Roadmap

Upcoming Features for v0.1

  • Regression
  • Backpropagation
  • Datasets, I/O

Upcoming Features for v0.2

  • Softmax layer
  • Classification
  • Integrate with MNIST for illustration

Features for v0.3+ (no particular order, nothing fixed)

  • Modularization
  • Convolutional neural networks
  • Recurrent neural networks
  • Backpropagation through time
  • Long-short term memory
  • ...