R.A.I.L: A Rust Artificial Intelligence Library
RAIL is designed to be a library for easily creating and training Neural Networks, akin to the Keras API. It aims to be fast and easy to use.
Dependencies
RAIL depends on arrayfire-rust, so before using RAIL make sure you have arrayfire installed.
A Simple XOR Problem
Solving the XOR Problem with Mold is super easy! Simply add the crate to your Cargo.toml:
= { = "https://github.com/nlsnightmare/rail" }
Then add this to your code
use Model;
use Dense;
use Activation;
Plans
As of now, RAIL is in a very early state, and under heavy development. The API will change a lot. So far, only Dense (aka fully connected) layers are supported, and batched SGD is the only way of training the network. However, there are plans to support:
- Convolutional Layers
- RNN Cells
- LSTM Cells
- Genetic Crossover
- ADAM optimizer
- More Activation functions
- More Error functions
- Documentation