Co-Designed Hardware and Algorithms Are Shrinking AI
Artificial intelligence (AI) is moving out of remote data centers and into smartphones, cameras, cars, robots, and other devices that process information locally. But running AI at the edge comes with an obvious problem: these low-power systems have limited computing power, battery capacity, and room for hardware. Even a seemingly simple task, such as determining which language a person is using, can consume resources that small devices cannot afford to waste.
Researchers from the University of Massachusetts Amherst have demonstrated a different approach that redesigns the algorithm and hardware together. Their system identified written languages with 95.24% accuracy while reducing computing resources by 90%, achieving the highest reported accuracy for a hyperdimensional computing implementation on an emerging hardware platform.
An overview of the approach (📷: Y. Huang et al.)
The work combines hyperdimensional computing (HDC) with analog in-memory computing hardware. HDC is a brain-inspired form of AI that represents information...
Copyright of this story solely belongs to hackster.io. To see the full text click HERE
