https://github.com/axect/noisy_candle

A Rust project showcasing regression on noisy data using machine learning libraries.
https://github.com/axect/noisy_candle

machine-learning rust

Last synced: 3 months ago
JSON representation

A Rust project showcasing regression on noisy data using machine learning libraries.

• Host: GitHub
• URL: https://github.com/axect/noisy_candle
• Owner: Axect
• License: mit
• Created: 2024-03-27T07:48:44.000Z (about 1 year ago)
• Default Branch: main
• Last Pushed: 2024-03-30T14:41:29.000Z (about 1 year ago)
• Last Synced: 2025-03-19T22:17:26.867Z (3 months ago)
• Topics: machine-learning, rust
• Language: Rust
• Homepage:
• Size: 1.04 MB
• Stars: 1
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# Noisy Regression with Candle

This project demonstrates how to perform regression on noisy data using various Rust libraries for machine learning. The main libraries used in this project are:

- `peroxide`: A library for data generation, preprocessing and visualization

- `candle`: A library for defining and training machine learning models

- `rayon`: A library for parallel processing

- `indicatif`: A library for progress tracking and visualization

## Key Features

1. Dataset Generation and Preprocessing

   - Generate noisy data using the `peroxide` library.

   - Split the data into train, validation, and test sets.

   - Scale the data using scalers such as `MinMaxScaler`, `StandardScaler`, and `RobustScaler`.

2. Model Definition and Training

   - Define an MLP (Multi-Layer Perceptron) model using the `candle` library.

   - Train the model using the Adam optimizer.

   - Track the training progress using the `indicatif` library.

3. Model Evaluation and Visualization

   - Evaluate the trained model on the test dataset and calculate the MSE.

   - Visualize the predictions using `plot` feature of `peroxide` library.

## Prerequisites

- Rust

- Python

- Required python libraries

  - `matplotlib`

  - `scienceplots` (optional)

## Usage

Just run the Rust code:

```sh

cargo run --release

```

## Results

- The training and validation losses are printed during the training process.

- The MSE on the test dataset is displayed after the model evaluation.

- The predicted results are visualized and saved to the `test_plot.png` file.

  ![test_plot.png](test_plot.png)

## LICENSE

This project is licensed under the MIT license.