Delay Line Reservoir Computer Structure

Delay Line Reservoir Computing (DLRC) is a recently introduced paradigm in the fields of machine learning and artificial neural networks (ANN). The Applied Physics Research Group (APHY) of the VUB developed a novel cascaded DLRC structure, aimed at real-time signal processing.

Our cascaded DLRC structure offers short-cycle supervised learning of complex nonlinear transformations from example and generalization to previously unseen inputs. 



Possible applications:

  • Nonlinear real-time audio processing
  • Timeseries forecasting
  • Noise suppression
  • Dropout compensation
  • General nonlinear filtering


Innovation Strengths / Competitive Advantages

  • Implementable as (embedded) DSP software or as IP block in an FPGA
  • Scalable and customizable
  • Simplified and fast training as compared to other ANN
  • Learning and execution platforms can be different
  • On-the-fly target function reconfiguration is possible

Market Opportunities / Industrial Target Group 

Our DLRC solution can be applied in any situation where sensor signals or input data needs to be discriminated, classified, or transformed via an a priori unknown or difficult to describe function or set of rules, defined by a (possibly limited) set of input-output examples.
Our target audience consists of the key players in the automotive, medical and industrial high-tech industries, challenged by complex, possibly real-time, signal processing requirements.

IP status



Reservoir Computing Device, European patent application nr. EP15200681.3 
Filed: 17 Dec. 2015.

Collaboration Details 


The Applied Physics Group of the Vrije Universiteit Brussel offers licensing, R&D partnerships and training to implement fully-tailored Delay Line Reservoir Computer solutions to your specific signal processing requirements.

Interested parties can contact

VUB TechTransfer
Hugo Loosvelt
[T]: +32 (0)2 629 38 65 or 22 07
[E]: hugo.loosvelt@vub.ac.be

Inventors 

Lars Keuninckx, Guy Van der Sande, Jan Danckaert,
Applied Physics Research Group (APHY),
Pleinlaan 2, Building F, 9th floor,
1050 Brussel.

[T]: +32 (0)2 629 38 46
[E]: lars.keuninckx@vub.ac.be
guy.van.der.sande@vub.ac.be

jan.danckaert@vub.ac.be

Keywords

signal processing, forecasting, reservoir computing, artificial neural networks.

Download a printable pdf here.