Ultra-Compact Energy-Efficient Neurocomputing Platform

Tech ID: 30445 / UC Case 2019-425-0

Brief Description

An energy, area, and speed efficient time-domain VMM circuit and neurotrophic processor architecture.

Background

Contemporary applications of computationally intensive artificial intelligence (AI) algorithms such as Deep/Reccurrent Neural Networks call for an efficient neuromorphic processor, especially for mobile/IoT devices. Within these deep neural networks and many other computationally-intensive data and signal processing systems, the vector-by-matrix multiplication (VMM) is the most common operation. Current digital approaches to the VMM task results in a relatively sparse design, which greatly diminishes the performance for memory access.

Description

Researchers at the University of California, Santa Barbara have created an energy, area, and speed efficient time-domain VMM circuit and neuromorphic processor architecture based on 3D-NAND flash memory devices targeting various AI applications. Efficient accelerators for emerging Non-Volatile Memory (NVM) devices such as compact flash memory devices are developed through the use of analog, rather than digital, computing.

Advantages


  • Smaller, faster, and energy efficient 
  • Potential for record-breaking storage efficiency, peak energy efficiency, and high computational throughout

Applications


  • Artificial Intelligence
  • Mobile & IoT devices
  • Robotics
  • Data-intensive or sensor-driven tasks
  • GPU based products

Patent Status

Patent Pending

Contact

Learn About UC TechAlerts - Save Searches and receive new technology matches

Inventors

Other Information

Keywords

electronics, indmicroelec, nueromorphic engineering, neurotrophic processor, artificial intelligence, neural networks, vector, internet of things

Categorized As