Quantum Hybrid Neural Decoding

In this project, the brain signals from a monkey are decoded by an LSTM with Variational Quantum Circuits. These results are compared against a classical LSTM to show that Quantum ML may be able to handle higher learning rates and learn more in initial epochs than the classical benchmark.

Acknowledgements

Thanks to Mike Vaiana @vaiana for guidance on neural decoding and for insights into the challenges in BCI today.

Thanks to NLB for the outstanding dataset:

Neural Latents Benchmark '21: Evaluating latent variable models of neural population activity¹

Key Files:

1. High-level Overview of Problem Statement & Findings --> HybridNeuralDecodingOverview.pdf (or for working links, use the Google slides version here).
2. Main Notebook with walkthrough and visuals --> Neural-Decoding.ipynb
3. Time Series Model Code including the QLSTM Cell --> TSModels.py

Main Results

The faster learning noted below can be attributed to a higher learning rate. When the Classical Model was trained with a comparably high learning rate, the loss was erratic and the model did not consistently converge - examples can be found in the Overview pdf.

Footnotes

1. Neural Latents Benchmark '21: Evaluating latent variable models of neural population activity, Felix Pei and Joel Ye and David M. Zoltowski and Anqi Wu and Raeed H. Chowdhury and Hansem Sohn and Joseph E. O’Doherty and Krishna V. Shenoy and Matthew T. Kaufman and Mark Churchland and Mehrdad Jazayeri and Lee E. Miller and Jonathan Pillow and Il Memming Park and Eva L. Dyer and Chethan Pandarinath; ARXIV: https://arxiv.org/abs/2109.04463 ↩