WINTER SCHOOL



"AI SIP: Hopfield Neural Network and Deep Learning"
December 3, 2024, Macau China
Introduction

Sponsor:  

Asia-Pacific Signal and Information Processing Association- APSIPA
Co-sponsors:  

APSIPA, Distinguished Lecturer Program, The University of Macau
Support:  

玖治科技（Ninenovo）
Theme:  

AI SIP: Hopfield Neural Network and Deep Learning
Lecturers:  

Prof. Mingyi He, Northwestern Polytechnical University

Prof. Ngai Fong Law, The Hong Kong Polytechnic University

Prof. Jen-Tzung Chien, National Yang Ming Chiao Tung University
Co-Chairs:  

Prof. Mingyi He, Northwestern Polytechnical University and APSIPA Education Program

Prof. Yuan Wu, University of Macau

Prof. Yuanman Li, Shenzhen University
Time:  

13:00-17:50 Dec. 03, 2024
Venue:  

Meeting Room 9, Galaxy International Convention Center


To promote APSIPA in the local region and expand education in signal and information technology, we propose organizing a winter school as a satellite event of APSIPA ASC 2024. The theme of the winter school will be "AI for SIP: Hopfield Neural Networks and Deep Learning." We plan to invite several renowned scholars in the field of AI to deliver a series of lectures.

Naturally physical signals (usually in analog form, such as sound, speech, images, videos, remote sensing, biomedical, X-ray imaging, etc.) are almost entirely converted into digital form, called "data", for transmission, storage, and processing in today's world. Recent major technical breakthroughs in AI (artificial intelligence) deep/machine learning in data processing (the main subject of signal and information processing - SIP) have had a huge impact on all scientific, technical, industrial, and service sectors. More excitingly, the 2024 Nobel Prize in Physics was recently awarded to John Hopfield and Geoffrey Hinton to recognize their fundamental discoveries and inventions in using artificial neural networks (ANN) for machine learning, which is also particularly relevant to the discipline of signal and information processing. To help researchers in the signal and information processing community, especially young graduates, to systematically and deeply understand their foundational contributions in AI machine/deep learning, as well as the impact and typical applications of deep learning in signal and information processing or data processing, the APSIPA Education Program organizes the APSIPA Winter School/Invited Lectures on "AI SIP: Hopfield Neural Network and Deep Learning" in Macau at APSIPA ASC 2024. This winter school courses are arranged as following four parts.

Part 1. Overview of Neural Network Al: Part 1 offers an brief overview of the evolution of neural networks/artificial intelligence (NN/AI), including Hopfield neural network (HNN), feedforward neural network, deep learning (with convolutions, skip-connections, transformers, etc) for signal and information processing. HNN laid the ground work for neural networks and machine learning, is highlighted for its foundational significance. The challenges on NN AI will also be discussed.
Part 2. Hopfield Neural Network Foundation for Machine Learning: Part 2 dives into the fundamentals of Hopfield neural network (HNN) with model, Lyapunov functions/stability, and applications to machine learning and SIP. From traditional HNN to modern HNN, from the Hopfield Lyapunov function for TSP to the Lyapunov function for TBP (Three Body Problem) discovered by deep learning in 2024, try to understand the intrinsic connection between HNN and deep learning and explore new insights to signal and information processing.
Part 3. Deep Learning for Image Forensics: Part 3 explores groundbreaking progress in machine and deep learning, emphasizing their remarkable discriminative abilities for tasks such as classification, recognition, and detection. Advanced models are examined in detail.
Part 4. Generative Modeling and Conversational Al: Part 4 focuses on the creative capabilities of modern machine learning models, such as content generation for articles, speeches, images, and videos. Highlighted technologies include transformer models and tools like ChatGPT, demonstrating their significant impact on generative modeling and conversational Al.


Content



13:00-14:00 by Mingyi He
1.1. From Nobel Prize Physics 2024 to the theme
1.2. Advance of Artificial Neural Networks
1.3. Typical Neural Networks/Machine Learning: MLP, HNN, RNN, DCNN, Skip-connection NN, Transformer, GPT, etc for SIP
1.4. The challenges on NN AI
1.5. Question and answer
14:00-15:00 by Mingyi He
2.1. Hopfield neural network (HNN) -Circuit model and analysis
2.2. Lyapunov function and dynamic system stability
2.3. Hopfield energy function and HNN stability
2.4. HNN for pattern recognition and optimization applications
2.5. Bolzmann machine for feature learning and generative models
2.6. Modern HNN in deep learning
2.7. Hopfield Lyapunov function for TSP and New Lyapunov function for TBP
2.8. Question and answer
15:30-16:30 by Bonnie Law
3.1. Introduction to Image forensics
3.2. Typical neural models for image forensics
3.3. Deep learning models for image forensics (source identification and forgery detection)
3.4. Advanced applications for DL (e.g., image tampering localization and open-set source identification)
3.5. Challenges
3.6. Discussion
16:30-17:30 by Jen-Tzung Chien
4.1. Background of spoken dialogue systems
4.2. Multimodal machine learning
4.3. Multilingual generative models
4.4. Pre-trained foundation models
4.5. Challenges and opportunities in comprehensive conversation system.
4.6. Discussion
17:30-17:50 All



Program



Northwestern Polytechnical University
Dr Mingyi He
"Hopfield Neural Network Fundamental for Machine Learning"

Mingyi He, Professor of Northwestern Polytechnical University (NPU), visiting professor of Adelaide Univ and Sydney Univ, Australia. His research interests focus on advanced machine vision and intelligent processing, including signal and image processing, computer vision, integrated image and graphics processing, hyper-spectral remote sensing, 3D information acquisition and processing, and neural network artificial intelligence. Prof. He published over 300 journal and conference papers, as well as 3 monographs (Neural Computing, Neural Network Signal Processing Systems, Digital Image Processing) etc. He is the recipient of 2012 IEEE CVPR best paper award, 2017 APSIPA best deep/machine learning paper award. He has obtained 11 scientific prizes, 3 teaching prizes from China (ministry and province government). He is a recipient of the government lifelong subsidy from the state council of China, 2012 Scientific Chinese, 2017 Baosteel outstanding teacher award. He also received certificates from the IEEE Signal Processing Society in 2014, APSIPA in 2019, the Chinese Institute of Electronics in 2018 and 2020, China Remote Sensing Committee in 2023. He has acted as the general chair or the TPC (co)chair for a number of international conferences. He had addressed a number of keynote talks or invited plenary talks. He is/was an Associate Editor or guest editor for IEEE TGRS, IEEE Jstars, APSIPA T-SIP, RS, JIG, SP etc. Professor He is CIE Fellow, CSIG Fellow, Vice president of APSIPA (21-26).


The Hong Kong Polytechnic University
Dr Ngai Fong (Bonnie) Law
"Deep Learning for Image forensics"

Bonnie is currently an associate professor at the Hong Kong Polytechnic University. Her research interests focus on signal and image processing, in particular image forensics which involves works on seam carving and source camera identification using deep learning. Throughout the years, she has engaged in various professional activities. For example, she served as the general co-chair of the 2017 International Conference on SP, Communication and Computing. She also held positions as the Treasurer of the IEEE HK Chapter of SP from 2009 to 2016 and as the secretary of APSIPA Headquarters since 2009. In 2011, she was awarded the Best paper award in the computing track at the IEEE Annual India Conference.



Kyushu National Yang Ming Chiao Tung University
Dr Jen-Tzung Chien
"Generative Modeling and Learning for Conversational AI"

Jen-Tzung Chien is currently the Lifetime Chair Professor in National Yang Ming Chiao Tung University, Hsinchu, Taiwan. He has authored more than 250 peer-reviewed articles in machine learning, deep learning, and Bayesian learning with applications on speech and natural language processing, and three books including Bayesian Speech and Language Processing, Cambridge University Press, 2015, Source Separation and Machine Learning, Academic Press, 2018, and Machine Learning for Speaker Recognition, Cambridge University Press, 2020. He was a Tutorial Speaker of AAAI, IJCAI, ACL, KDD, ICASSP, COLING and Interspeech. He received the Best Paper Award in IEEE Workshop on Automatic Speech Recognition and Understanding in 2011, and IEEE International Workshop on Machine Learning for Signal Processing in 2023.



SPONSORSHIP