Sparse Autodiff: Efficient Sparse Matrix and Vector Support in Needle

Efficient sparse matrix and vector support with automatic differentiation in Needle framework

About Needle

Needle (Necessary Elements of Deep Learning) is a mini deep learning framework proposed by Carnegie Mellon University (CMU) for teaching purposes. It includes the basic components of modern deep learning frameworks, making it an excellent educational tool for understanding the inner workings of frameworks like PyTorch and TensorFlow.

Repository

The Needle framework is available on GitHub: iwzbi/needle

This project extends Needle by adding efficient sparse matrix and vector support with automatic differentiation capabilities, enabling sparse tensor operations within the Needle framework.

Project Demo Video

Project demonstration video showcasing the Sparse Autodiff implementation and features.

Jupyter Notebook Analysis