NNRS - Rust Neural Network Library

NNRS is a Rust library for creating and working with feedforward neural networks. It provides a set of tools for building and manipulating neural networks, including creating nodes and edges, setting inputs, and firing the network to generate outputs.

Note: this library is still in development, and is not yet ready for use.

Installation

To use NNRS, simply add it as a dependency to your Rust project with Cargo:

cargo add nnrs

Usage

For the most up-to-date usage information, clone the repository and run cargo doc --open, or view the documentation on docs.rs.

Limitations

At this moment, NNRS does not include training functionality. You can use this library to generate outputs from pre-trained networks.

Roadmap

• Basic Parts
• Generating Outputs
• Serialization
• Training (NEAT?)

License

NNRS is licensed under the AGPLv3 license. See the LICENSE file for more information.

lib.rs:

nnrs

A simple, minimal neural network library written in Rust. No training included

Example:

use nnrs::{network::Network, node::Node, edge::Edge, layer::LayerID, activationfn::ActivationFn};

let mut network = Network::create(1, 1, ActivationFn::Linear).unwrap();
let mut output: Vec<f64> = Vec::new();

let layer_id = network.add_layer();

let input_node_id = network.input_node_ids().pop().unwrap();
let hidden_node_id = Node::create(&mut network, layer_id, 0.2).unwrap();
let output_node_id = network.output_node_ids().pop().unwrap();

let edge_input_to_hidden = Edge::create(&mut network, input_node_id, hidden_node_id, 1.3).unwrap();
let edge_hidden_to_output = Edge::create(&mut network, hidden_node_id, output_node_id, 1.5).unwrap();

network.fire(vec![0.8], &mut output).unwrap();

assert_eq!(output, vec![((0.8 * 1.3) + 0.2) * 1.5]);