Brain-Inspired Algorithms, Architectures and Circuits

Generated by Stable Diffusion Text-to-Image Model

The brain is the perfect place to look for inspiration to build more efficient computers. Our goal in the UCSC Neuromorphic Computing Group led by Assistant Prof. Jason Eshraghian is to understand the computational principles that underpin the brain, and use them to engineer more efficient systems that can adapt to ever-changing environments. We develop algorithms that can learn, and low-power architectures and circuits that harness exotic device technologies. Our work sits at the intersection of neuroscience, deep learning, and VLSI design.

We maintain snnTorch, a deep learning Python library that enables gradient-based optimization of spiking neural network and has been downloaded over 100,000 times. We also developed the MatMul-free LM, SpikeGPT amongst a bunch of other cool stuff.

Photo credit: Jonathan Morris

Collaborations and Applications

Our research is actively used across domains in both research and applied settings. Cardiologists apply our neuromorphic vision algorithms to assess the risk of heart failure. Neurologists have successfully used spiking neural networks to forecast the early onset of seizures. Outside of the hospital, the same models have been used to track space junk orbiting the Earth to avoid high-risk missions, and for forecasting power loads from renewable energy sources in rural areas.

We’re open to collaborations and sponsorship opportunities. Feel free to get in touch at jeshragh@ucsc.edu.

Interested in joining us? Check out current openings here

Sponsors