Graph Neural Networks for Classical and Quantum Channel Coding

Explore a groundbreaking approach to channel decoding using graph neural networks (GNNs) in this 34-minute talk by Sebastian Cammerer from Nvidia. Delve into a fully differentiable GNN-based architecture that learns generalized message passing algorithms over given graph structures representing forward error correction codes. Discover how this innovative method overcomes scalability issues and the curse of dimensionality prevalent in other deep learning-based decoding approaches. Compare the performance of this proposed decoder against state-of-the-art conventional and deep learning-based channel decoding techniques. Examine the extension of this GNN architecture to quantum low-density parity-check (LDPC) codes, integrating classical belief propagation with GNN layers on the same sparse decoding graph. Learn how careful training processes can significantly lower error floors, advancing both classical and quantum channel coding techniques.

Graph Neural Networks for Classical and Quantum Channel Coding