Introduction

This course is mainly a new professional elective course for electrical undergraduates such as Computer Science and Technology, Artificial Intelligence, Internet-of-things Engineering and Microelectronics Science and Engineering. The course takes the design methodology of Artificial Intelligent (AI) Processors as the main line, and focuses on the topics of intelligent algorithm, intelligent programming framework, intelligent chip architecture, intelligent programming language, AI chip cutting-edge technology, innovative computing paradigm and hardware implementation. The course pays attention to the combination of theory and practice, pays attention to practicability, improves students' theoretical level and practical ability of AI SoC design. The objective of the course is to master the architecture and design methodology of AI SoC, and enable students to understand the complete software and hardware technology system of AI SoC and system.

This course is mainly carried out in the form of a combination of theoretical teaching and large assignments. The teaching contents include four parts: 1) introduction to neural network and deep learning；2）hardware architecture of deep learning based on CNN and RNN algorithms; 3) Principle and hardware of spike neural network; 4) Cutting-edge technologies of AI SoCs.

The course has 32 class hours in total. The practice part requires students to complete the hardware implementation of an intelligent algorithm according to the course content, and complete the verification of hardware circuit through FPGA development. This course adopts a combination of homework and oral defense to finish the examination.

Contents

Module 1 – Introduction to AI processors

Module 2 – Basis of Neural Network and Deep Learning

Module 3 – Operational Principle of Deep Learning Hardware Accelerator

Module 4 – Hardware Architecture of DNN processors

Module 5 – Principle of Spike Neural Network

Module 6 – Hardware Architecture of SNN processors

Module 7 – Mixed-Signal Computing Paradigms and Chip Implementation

Module 8 - Cutting-edge technologies of AI SoCs

Reference textbooks

Yunji Chen, et al.,”Intelligent Computing Systems”，China Machine Press，April, 2020

Nan Zheng et al.，“Learning in Energy-Efficient Neuromorphic Computing: Algorithm and Architecture Co-Design”，Wiley-IEEE Press，Dec. 2019

Chenxiong Zhang，”AI chips: Cutting-Edge Technologies and Innovative Future”， Posts and telicommunications Press，April, 2021

Vivienne Sze et al.，” Efficient Processing of Deep Neural Networks”，Morgan &Claypool publishers，April, 2020

Courses for students

Undergraduate students

Master students

Spring 2022， Undergraduate students， 32 hours

1 – Introduction to AI processors (PDF)

2 – Basis of Neural Network and Deep Learning (PDF)

3 – Operational Principle of Deep Learning Hardware Accelerator (PDF)

4 – Hardware Architecture of DNN processors (PDF)

5 – Principle of Spike Neural Network (PDF)

6 – Hardware Architecture of SNN processors (PDF)

7 – Mixed-Signal Computing Paradigms and Chip Implementation (PDF)

8 - Cutting-edge technologies of AI SoCs (PDF)