Neural Networks (in Rust)

A convolutional neural network with enterprise skillz built in Rust with no dependancies supporting hardware acceleration

Used in part of a 5 part article on LinkedIn (https://www.linkedin.com/in/samcolak) to understand how to develop Neural Networks using a stable and solid foundation

Written by Samuel Colak (sam@samcolak.com)

Overview

• /library : Multimodal Convolutional Neural Network Library for rust
• /pt1 : Part 1 of developing a neural network
• /pt2 : Part 2
• /pt3 : Part 3
• /pt4 : Part 4
• /pt5 : Active workload to test library functionality

Distribution

The is distributed under the GPL-3 license

Any derived works released must be open-sourced - This is to benefit the community and foster development in the ML / NeuralNet space

Coverage

Coverage checks are wired via cargo-llvm-cov for both active crates:

• library/neuralnet
• pt5/neuron

Run locally:

cargo install cargo-llvm-cov
bash scripts/coverage.sh

Optional local gate (line coverage %):

COVERAGE_MIN_LINES=35 bash scripts/coverage.sh

CI workflow:

• .github/workflows/coverage.yml
• uploads LCOV artifacts for both crates as coverage/*.lcov