NEMS resonator modes combine to boost computing power
Dynamical Diversity for Reservoir Computing in Reconfigurable Nanomechanics
Emerging TechnologiesNeural and Evolutionary Computing
Summary
Processing information using physical devices requires capturing complex patterns in signals. The authors showed that using two interacting vibrational modes of a tiny mechanical device instead of one can dramatically improve how well the device can handle complicated tasks. They do this by feeding the input in different ways to these modes and then mixing the responses, which creates richer data for the computer to learn from. Their tests showed much lower errors on tough tasks compared to using a single mode or separate modes alone. This approach unlocks more computing ability from a single small device without needing extra parts.
What this means in practice
- •For embedded systems engineers: Improve small-scale physical computing devices by combining resonance modes to enhance nonlinear signal processing without adding hardware complexity.
- •For robotics control teams: Deploy compact mechanical neural networks with expanded computational capabilities for handling nonlinear control tasks in robots and autonomous systems.
Authors
Humayun Ahmed, Inês S. Garcia, Filipa C. Mota, Muhammad Asim, Ghaith Mahook, Suleyman Celik, Yunus Selcuk, Filipe S. Alves, Alper Demir, M. Selim Hanay
Abstract
Physical reservoir computing uses nonlinear dynamics and a trained linear readout to process information. Nanoelectromechanical (NEMS) resonators combine geometric Duffing nonlinearity with fading memory, but most electromechanical implementations use a single resonance mode. Here, we demonstrate reservoir computing with two interacting modes of a single NEMS resonator measured through one readout port. We introduce dynamical diversity through complementary modal drive settings: the same input sequence is replayed under different allocations of drive amplitude between the modes, and the responses are concatenated into a single feature matrix. This multiplexing expands the representation available to the readout without additional devices or training of internal parameters. On NARMA-2, it reduces variance-normalized test error more than 28-fold relative to single-mode operation and more than threefold relative to the best individual two-mode setting. Linear memory-capacity measurements show that accessible recall spans only a few symbols at the tested symbol duration. Its rapid decline with delay, consistent with mechanical dissipation, accompanies rising NARMA error and the eventual loss of multiplexing gains at higher orders. We also use electrical feedthrough as an internal reference for assessing the computational contribution of the NEMS response. Separate linear readouts are trained on feedthrough features and features derived from the measured NEMS response, using the same recordings and matched drive settings and processing. On challenging nonlinear mapping tasks, the multiplexed NEMS features yield substantially lower errors than the feedthrough features. These results demonstrate how dynamical diversity through variations in modal drive amplitudes expands the computational capability of a single multimode NEMS resonator.