1 h 1 min
Seminar, Hosted by MIT.nano Seminar Series
Neuromorphic Computing via Integration of Diffusive and Drift Memristors
University of Southern California
Recording
Abstract
Memristors fall into two classes, diffusive and drift, distinguished by how they reset. Diffusive memristors reset through spontaneous ion diffusion at zero bias, behaviour that resembles biological ionic dynamics and suits temporal information processing. Drift memristors rely on field-driven ion migration, giving stable, tunable analogue states for long-term storage. J. Joshua Yang shows how integrating the two types enables compact, energy-efficient and biorealistic neuromorphic hardware, covering device engineering, array integration and application demonstrations for machine-learning and AI hardware.
Topics
- memristors
- neuromorphic computing
- AI hardware, broader topic: hardware
- Diffusive memristors, broader topic: memristors
- Drift Memristors, broader topic: memristors
- ionic dynamics, broader topic: dynamics
- Temporal information processing, broader topic: information processing
- ion diffusion, broader topic: diffusion
- field-driven ion migration, broader topic: migration
- analogue states, search World Wide
- machine-learning and AI hardware, broader topic: hardware