grad_rs

• Cargo.lock
• Cargo.toml
• README.txt
• data
• src

README

GRAD_RS
by anders

A CPU based simple gradient descent neural network training program.
Currently just a conceptual implementation, it does not store the neural network for later use.
grad_rs is set up to automatically parse and interpret CSV files.
Inspired by https://github.com/karpathy/micrograd.

Included are also some example datasets taken from https://archive.ics.uci.edu/datasets.


Building:
cargo build

Usage:
grad_rs [OPTIONS] \ 
--training-data-path <TRAINING_DATA_PATH> \ 
--learning-rate <LEARNING_RATE> \  
--stopping-loss-percentage <STOPPING_LOSS_PERCENTAGE>

Options:
  -t, --training-data-path <TRAINING_DATA_PATH>
  -l, --learning-rate <LEARNING_RATE>
  -s, --stopping-loss-percentage <STOPPING_LOSS_PERCENTAGE> (stop the training when certain loss score is reached)
  -c, --class-column-index <CLASS_COLUMN_INDEX>              [default: -1]
  -h, --help                                                 Print help
  -V, --version

Example usage:
./target/release/grad_rs -t data/car.data.txt -l 0.0005 -s 0.1

Download