Advances in Neuromorphic Computing: Models, Hardware, and Optimization 

ICASSP 2026 Tutorial — April 5, 2026

About This Tutorial

Modern AI is transforming the world — but at a steep energy cost. This tutorial explores neuromorphic computing, a brain-inspired computing paradigm aimed at dramatically improving the intelligence-to-watt ratio of AI systems.

Spanning three hours, this tutorial brings together experts from Northeastern University and Cornell University to cover the full neuromorphic computing stack: from computational models and specialized hardware to training and optimization algorithms.

What You Will Learn

• How spiking neural networks exploit dynamic sparsity and event-driven communication for energy efficiency

• How neuromorphic computing units can implement intra-token and inter-token processing

• How to design neuromorphic state-space models and neuromorphic transformers

• The principles of in-memory computing and how co-locating memory and computation overcomes the von Neumann bottleneck

• Local learning rules as biologically plausible, memory-efficient alternatives to backpropagation

• The latest advances in neuromorphic hardware platforms

• Practical optimization and training techniques for neuromorphic systems

Tutorial Structure and Slides

1. Part I: Models (Osvaldo Simeone) - 1 hour

2. Part II: Hardware (Bipin Rajendran) - 1 hour

3. Part III: Optimization and Training (Tianyi Chen) - 1 hour

Presenters

• Osvaldo Simeone and Bipin Rajendran — Northeastern University London

• Tianyi Chen — Cornell University