Research Manuscript: Memory - The Workhorse of Machine Learning
Event TypeResearch Manuscript
Hosted in Virtual Platform
DescriptionMemory management and optimization are key considerations for machine learning system design. Memory access dominates power consumption and limits the size of applications that can be run on a system. This session takes on this critical topic with four exciting papers on how to use or implement memory more efficiently for machine learning. Two papers in this session use algorithmic techniques to compress memory usage and the following two papers implement efficient and reliable neural network systems based on ReRAM and beyond.