CaltechAUTHORS
  A Caltech Library Service

Deep Multi-State Dynamic Recurrent Neural Networks Operating on Wavelet Based Neural Features for Robust Brain Machine Interfaces

Haghi, Benyamin and Kellis, Spencer and Shah, Sahil and Ashok, Maitreyi and Bashford, Luke and Kramer, Daniel and Lee, Brian and Liu, Charles and Andersen, Richard A. and Emami, Azita (2019) Deep Multi-State Dynamic Recurrent Neural Networks Operating on Wavelet Based Neural Features for Robust Brain Machine Interfaces. In: 33rd Conference on Neural Information Processing Systems. Neural Information Processing Systems Foundation, Inc. , Art. No. 9594. https://resolver.caltech.edu/CaltechAUTHORS:20190724-154847448

[img] PDF - Published Version
See Usage Policy.

1076Kb
[img] PDF - Submitted Version
See Usage Policy.

3968Kb
[img] Archive (ZIP) - Supplemental Material
See Usage Policy.

3083Kb

Use this Persistent URL to link to this item: https://resolver.caltech.edu/CaltechAUTHORS:20190724-154847448

Abstract

We present a new deep multi-state Dynamic Recurrent Neural Network (DRNN) architecture for Brain Machine Interface (BMI) applications. Our DRNN is used to predict Cartesian representation of a computer cursor movement kinematics from open-loop neural data recorded from the posterior parietal cortex (PPC) of a human subject in a BMI system. We design the algorithm to achieve a reasonable trade-off between performance and robustness, and we constrain memory usage in favor of future hardware implementation. We feed the predictions of the network back to the input to improve prediction performance and robustness. We apply a scheduled sampling approach to the model in order to solve a statistical distribution mismatch between the ground truth and predictions. Additionally, we configure a small DRNN to operate with a short history of input, reducing the required buffering of input data and number of memory accesses. This configuration lowers the expected power consumption in a neural network accelerator. Operating on wavelet-based neural features, we show that the average performance of DRNN surpasses other state-of-the-art methods in the literature on both single- and multi-day data recorded over 43 days. Results show that multi-state DRNN has the potential to model the nonlinear relationships between the neural data and kinematics for robust BMIs.


Item Type:Book Section
Related URLs:
URLURL TypeDescription
https://papers.nips.cc/paper/9594-deep-multi-state-dynamic-recurrent-neural-networks-operating-on-wavelet-based-neural-features-for-robust-brain-machine-interfacesPublisherArticle
https://doi.org/10.1101/710327DOIDiscussion Paper
ORCID:
AuthorORCID
Haghi, Benyamin0000-0002-4839-7647
Kellis, Spencer0000-0002-5158-1058
Bashford, Luke0000-0003-4391-2491
Kramer, Daniel0000-0003-4551-2977
Andersen, Richard A.0000-0002-7947-0472
Emami, Azita0000-0003-2608-9691
Additional Information:© 2019 Neural Information Processing Systems Foundation, Inc. We thank Tianqiao and Chrissy (T&C) Chen Institute for Neuroscience at California Institute of Technology (Caltech) for supporting this IRB approved research. We also thank Dr. Erin Burkett for reviewing this manuscript.
Group:Tianqiao and Chrissy Chen Institute for Neuroscience
Funders:
Funding AgencyGrant Number
Tianqiao and Chrissy Chen Institute for NeuroscienceUNSPECIFIED
Record Number:CaltechAUTHORS:20190724-154847448
Persistent URL:https://resolver.caltech.edu/CaltechAUTHORS:20190724-154847448
Usage Policy:No commercial reproduction, distribution, display or performance rights in this work are provided.
ID Code:97394
Collection:CaltechAUTHORS
Deposited By: Tony Diaz
Deposited On:24 Jul 2019 22:56
Last Modified:09 Jul 2020 21:37

Repository Staff Only: item control page