DFG SPP 2353: Dynamics-Informed Reservoir Computing (DIRC)

Apr 1, 2023 · 2 min read
Manish Yadav

This project is part of the DFG Priority Program SPP 2353: Daring More Intelligence – Design Assistants in Mechanics and Dynamics, funded by the German Research Foundation (DFG).

Team

Project Overview

The DIRC (Dynamics-Informed Reservoir Computing) project aims to develop novel design assistants with regard to complex dynamic loads in structural dynamics using dynamically integrated methods of machine learning. The core focus is on reservoir computing approaches for fast and green machine learning.

DIRC Project schematic
Figure 1: Schematic of the current approach for ‘Desiging for Dynamics’ versus our proposed approach with RC-based design assitant DIRC that increases efficiency.

Research Goals

  • Dynamics-informed machine learning: Building design assistants based on dynamics-informed ML approaches specifically tailored for structural mechanics applications
  • Reservoir Computing: Developing fast and energy-efficient computing paradigms using reservoir networks
  • Performance-dependent network evolution: Understanding how network structures evolve to optimize computational performance
  • Structure-function relationships: Investigating the relationship between network topology and computational capabilities

  • M. Yadav, S. Sinha, M. Stender, “Evolution beats random chance: Performance-dependent network evolution for enhanced computational capacity,” Phys. Rev. E 111, 014320 (2025)

  • M. Yadav and M. Stender, “Task-specific node pruning enhances computational efficiency of reservoir computing networks,” Chaos 35, 083123 (2025)

Funding

This project is funded by the German Research Foundation (DFG) as part of the Priority Program SPP 2353.

Design Assistants DFG Reservoir Computing Machine Learning Structural Mechanics
Dr. Manish Yadav
Authors
Dr. Manish Yadav
Physicist & ML Researcher, Postdoctoral Scientist