Pruning deep neural networks generates a sparse, bio-inspired nonlinear controller for insect flight

PLoS Comput Biol. 2022 Sep 27;18(9):e1010512. doi: 10.1371/journal.pcbi.1010512. eCollection 2022 Sep.

Abstract

Insect flight is a strongly nonlinear and actuated dynamical system. As such, strategies for understanding its control have typically relied on either model-based methods or linearizations thereof. Here we develop a framework that combines model predictive control on an established flight dynamics model and deep neural networks (DNN) to create an efficient method for solving the inverse problem of flight control. We turn to natural systems for inspiration since they inherently demonstrate network pruning with the consequence of yielding more efficient networks for a specific set of tasks. This bio-inspired approach allows us to leverage network pruning to optimally sparsify a DNN architecture in order to perform flight tasks with as few neural connections as possible, however, there are limits to sparsification. Specifically, as the number of connections falls below a critical threshold, flight performance drops considerably. We develop sparsification paradigms and explore their limits for control tasks. Monte Carlo simulations also quantify the statistical distribution of network weights during pruning given initial random weights of the DNNs. We demonstrate that on average, the network can be pruned to retain a small amount of original network weights and still perform comparably to its fully-connected counterpart. The relative number of remaining weights, however, is highly dependent on the initial architecture and size of the network. Overall, this work shows that sparsely connected DNNs are capable of predicting the forces required to follow flight trajectories. Additionally, sparsification has sharp performance limits.

Publication types

  • Research Support, Non-U.S. Gov't

MeSH terms

  • Animals
  • Insecta*
  • Neural Networks, Computer*

Grants and funding

This work was supported by AFOSR Grant (Nature-Inspired Flight Technology and Ideas, FA9550-14-1-0398) to TLD and the Komen Endowed Chair to TD. CS was supported in part by the Washington Research Foundation and by a Data Science Environments project award from the Gordon and Betty Moore Foundation (Award \#2013-10-29) and the Alfred P. Sloan Foundation (Award \#3835) to the University of Washington eScience Institute. JB was supported in part by the National Science Foundation Graduate Research Fellowship Program. JNK and TD acknowledge support from the Air Force Office of Scientific Research (FA9550-19-1-0386). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Nature-Inspired Flight Technology and Ideas: http://washington.edu/#:~:text=Technologies%20and%20Ideas-,The%20Air%20Force%20Center%20of%20Excellence%20on%20Nature%2DInspired%20Flight,of%20flight%20in%20complex%20environments. Washington Research Foundation: https://www.wrfseattle.org/ Gordon and Betty Moore Foundation: https://www.moore.org/ Alfred P. Sloan Foundation: https://sloan.org/ NSF: https://www.nsf.gov/ Air Force Office of Scientific Research: https://www.afrl.af.mil/AFOSR/.