A functional, type-safe library for deep machine learning in Haskell.
Status: Early development stage. APIs may change significantly.
Deep ML aims to provide a strongly-typed, composable framework for building and training deep learning models in Haskell. The project emphasizes type safety, mathematical correctness, and functional programming principles.
Automatic differentiation library with reverse-mode backpropagation.
- Reverse-mode automatic differentiation (backpropagation)
- Support for higher-order derivatives
- Type-safe gradient computation
- Integration with NumHask for polymorphic numeric operations
- Flexible representations including profunctor and Van Laarhoven encodings
Symbolic expression library for debugging and mathematical verification.
- Symbolic representation of mathematical expressions
- Expression simplification and manipulation
- Useful for debugging automatic differentiation computations
- Human-readable output for complex mathematical operations
- Expression graph visualization using Graphviz
# Install from Hackage
cabal install simple-expr
cabal install inf-backpropOr add to your project's *.cabal file
build-depends: simple-expr, inf-backprop Stakage users can add the packages to their stack.yaml
dependencies: simple-expr, inf-backpropSee inf-backprop tutorial and simple-expr tutorial for step-by-step guides to get started with each package.
- Core automatic differentiation engine
- Basic numeric type support
- Tensor operations
- GPU acceleration support
- Neural network layers
- Optimization algorithms
- Pre-trained model zoo
This project is in early development and we're actively seeking feedback.
Please feel free to:
- Report bugs and issues
- Suggest new features
- Improve documentation
- ad - Automatic differentiation
- backprop - Heterogeneous automatic differentiation
- grenade - Dependently typed neural networks
- hasktorch - Haskell bindings to PyTorch
BSD 3-Clause License. See the LICENSE file for details.