Single chip photonic deep neural network with accelerated training · HackerPapers discussions

Single chip photonic deep neural network with accelerated training — mentioned by programjames

Tue, 27 May 2025 00:48:27 GMT

There is, and people have trained purely optical neural networks: https://arxiv.org/abs/2208.01623 The real issue is trying to backpropagate those nonlinear optics. You need a second nonlinear optical component that matches the derivative of the first nonlinear optical component. In the paper above, they approximate the derivative by slightly changing the parameters, but that means the training time scales linearly with the number of parameters in each layer. Note: the authors claim it takes O(sqrt N) time, but they're forgetting that the learning rate mu = o(1/sqrt N) if you want to converge…

Paper and discussion history. A mention is not an endorsement.

Single chip photonic deep neural network with accelerated training — mentioned by adrian_b

Wed, 15 Jan 2025 10:18:16 GMT

The preprint of the research paper: https://arxiv.org/abs/2208.01623

Paper and discussion history. A mention is not an endorsement.

Single chip photonic deep neural network with accelerated training — mentioned by RansomStark

Tue, 03 Dec 2024 20:26:40 GMT

Arxiv link: https://arxiv.org/abs/2208.01623

Paper and discussion history. A mention is not an endorsement.