Kolmogorov–Arnold Network (KAN) in Rust

This project is a Rust implementation of a Kolmogorov–Arnold Network (KAN) neural network. The KAN network is a type of feedforward neural network that uses a spline activation function to approximate any continuous function. The network is trained using backpropagation and gradient descent to minimize the loss function. The project includes a library for building and training the network, as well as an example application that demonstrates how to use the network to solve a regression problem.

Description

• src/bin/kan.rs: The main entry point of the application.
• src/data_structures: Contains various data structures used in the project like KANLayer, layer, matrix, spline, and vector.
• src/lib.rs: The library file.
• src/network: Contains the network implementation.
• src/tests: Contains the unit tests for the various components of the project.
• src/utils: Contains utility functions and modules like activations, is_close_enough, and loss_functions.
• model and model.json: These files are related to the model used in the project.

How to Run

To run the project, use the following command:

cargo run

How to Test

To run the tests, use the following command:

cargo test

License

This project is licensed under the MIT License.

Author

Andres Caicedo

Acknowledgements

• Rust Programming Language
• Kolmogorov–Arnold Networks

Future Work

• Implement more advanced features like dropout and batch normalization.
• Optimize the code for better performance.
• Explore different applications of the KAN network.
• Add more unit tests and integration tests.
• Create a more user-friendly interface for training and using the network.
• Implement a GUI for visualizing the network and its results.

Contributions

Contributions are welcome! Please feel free to submit pull requests or open issues.

• Add more activation functions: Implement more activation functions like ReLU, sigmoid, and tanh.
• Implement different loss functions: Implement different loss functions like mean squared error, cross-entropy, and hinge loss.
• Add support for different datasets: Add support for different datasets like MNIST, CIFAR-10, and ImageNet.
• Implement different optimization algorithms: Implement different optimization algorithms like Adam, SGD, and RMSprop.
• Improve the documentation: Improve the documentation of the code and the project.
• Add more examples: Add more examples of how to use the KAN network.
• Create a GUI: Create a GUI for visualizing the network and its results.

This project is still under development, but I hope it will be a useful resource for anyone interested in learning about KAN neural networks and implementing them in Rust.