Artificial intelligence (AI) has many applications in today's society, including robots intelligence, traffic control, data analytics, image recognition, and speech understanding. The growing size and complexity of AI Algorithms require high-performance computation and memory resources. Application-specific hardware and emerging devices/systems are needed to achieve orders of magnitude improvement in performance and energy efficiency of AI algorithms.
The goal of this project is to research and develop algorithms, architectures, and digital implementations of spiking neural networks technology that will have a big impact on IoTs, robotics, as well as prosthetic devices. The approach is to develop a robust, real-time, and low-power neuro-inspired solutions targeted for full-custom system-on-chip integration and featuring the followings: (1) adaptive configuration method which enables reconfiguration of different network parameters (spike weights, routing, hidden layers, topology, etc.), (2) a mixture of different deep neural network topologies, (3) an efficient fault-tolerant spike routing algorithm, and (4) online learning mechanism. To demonstrate the performance of our processors/systems, an FPGA implementation shall be developed. Besides, a VLSI implementation shall be established.