Phase Change Memory

Phase-change memory (PCM) is being actively explored as a foundation for energy-efficient in-memory computing, aiming to overcome limitations of traditional von Neumann architectures. Current research focuses on integrating PCM with various neural network architectures, including Bayesian neural networks and spiking neural networks, to enable efficient and robust machine learning applications like image classification and robotic control. This work emphasizes techniques like learning-to-learn and continual learning to improve adaptation and reduce the need for extensive retraining, while also addressing challenges such as device stochasticity and write endurance. The ultimate goal is to create low-power, high-performance AI systems suitable for edge devices and resource-constrained environments.