Mod-DeepESN: Modular Deep Echo State Network

TL;DR

Mod-DeepESN introduces a modular deep echo state network with intrinsic plasticity to handle complex, multi-scale temporal tasks, outperforming existing models in time series prediction.

Contribution

It presents a novel modular deep echo state network architecture infused with intrinsic plasticity for improved complex temporal task performance.

Findings

Outperforms state-of-the-art in time series prediction

Effective for multi-scale, complex temporal tasks

Demonstrates superior adaptability over traditional echo state networks

Abstract

Neuro-inspired recurrent neural network algorithms, such as echo state networks, are computationally lightweight and thereby map well onto untethered devices. The baseline echo state network algorithms are shown to be efficient in solving small-scale spatio-temporal problems. However, they underperform for complex tasks that are characterized by multi-scale structures. In this research, an intrinsic plasticity-infused modular deep echo state network architecture is proposed to solve complex and multiple timescale temporal tasks. It outperforms state-of-the-art for time series prediction tasks.