Integer Echo State Networks: Hyperdimensional Reservoir Computing

06/01/2017
by   Denis Kleyko, et al.
0

We propose an integer approximation of Echo State Networks (ESN) based on the mathematics of hyperdimensional computing. The reservoir of the proposed Integer Echo State Network (intESN) contains only n-bits integers and replaces the recurrent matrix multiply with an efficient cyclic shift operation. Such an architecture results in dramatic improvements in memory footprint and computational efficiency, with minimal performance loss. Our architecture naturally supports the usage of the trained reservoir in symbolic processing tasks of analogy making and logical inference.

READ FULL TEXT

page 1

page 2

page 3

page 4

research
06/12/2020

Reservoir Computing meets Recurrent Kernels and Structured Transforms

Reservoir Computing is a class of simple yet efficient Recurrent Neural ...
research
01/08/2023

Stochastic Reservoir Calculations

Prabhu (1958) obtained the stationary distribution of storage level Z_t ...
research
11/29/2022

Optimizing time-shifts for reservoir computing using a rank-revealing QR algorithm

Reservoir computing is a recurrent neural network paradigm in which only...
research
09/24/2019

Reservoir Topology in Deep Echo State Networks

Deep Echo State Networks (DeepESNs) recently extended the applicability ...
research
06/06/2022

Reservoir Computing in robotics: a review

Reservoir Computing is a relatively new framework created to allow the u...
research
07/25/2018

Pre-trainable Reservoir Computing with Recursive Neural Gas

Echo State Networks (ESN) are a class of Recurrent Neural Networks (RNN)...
research
07/17/2023

Reducing hyperparameter dependence by external timescale tailoring

Task specific hyperparameter tuning in reservoir computing is an open is...

Please sign up or login with your details

Forgot password? Click here to reset