Seminar on "Two-Model Synergetic Learning System Optimization with Maxwell’s Demon Technique"
IEEE Northern Jersey Section SMC Chapter Seminar
Two-Model Synergetic Learning System Optimization with Maxwell’s Demon Technique
Ping Guo, Ph.D. & Professor
School of Systems Science
Beijing Normal University
Venue: ECEC 202, New Jersey Institute of Technology, Newark, NJ, USA
Time: 10-11am, Dec. 4, 2023 (Monday)
Online: https://njit.webex.com/join/zhou
Abstract
A two-model Synergetic Learning Systems (2MSLS) is a special case of the synergetic learning systems. It has two models (subsystems), i.e., a reduction model and an evolution model, which are governed by neural partial differential equations. The information ``particles'' diffusion process is described by the evolution model that is equivalent to a generative model. The particles condensate process is described by a reduction model that is equivalent to an inference model. In order to optimize 2MSLS, we propose a novel optimization scheme, named a Maxwell's demon technique (MDT). MDT is applied to decrease system entropy. It is an integrated technique that includes Bayesian pseudoinverse learners, pseudoinverse learning algorithm for autoencoders, and probabilistic principal component analysis. Theoretical analysis shows that our proposed 2MSLS has stronger interpretability with statistical physics image. MDT has high efficiency in optimization compared with Monte Carlo Markov Chain sampling methods. This work is’s significance lies in that it presents a clear view of interpretable deep neural networks with statistical physics perspective, and paves the way to physical artificial intelligence.
Bio-sketch
Ping Guo, Professor, IEEE senior member, CCF senior member, School of systems science, Beijing Normal University; Chair of IEEE CIS Beijing Chapter (2015-2016). His research interests include computational intelligence theory and its applications in pattern recognition, image processing, software reliability engineering, and astronomical data processing. He has published more than 360 papers, held 6 patents, and authored two books: “Computational intelligence in software reliability engineering”, and “Image semantic analysis.” He received 2012 Beijing municipal government award of science and technology (third rank) entitled "regularization method and its application". Professor Guo received his Master's degree in optics from the Department of Physics, Peking University, and received his Ph.D. degree from the Department of Computer Science and Engineering, Chinese University Hong Kong.
Date and Time
Location
Hosts
Registration
- Date: 04 Dec 2023
- Time: 09:59 AM to 11:15 AM
- All times are (UTC-05:00) Eastern Time (US & Canada)
-
Add Event to Calendar
- 323 MLK Blvd.
- Newark, New Jersey
- United States 07102
- Starts 26 November 2023 09:59 PM
- Ends 03 December 2023 09:59 PM
- All times are (UTC-05:00) Eastern Time (US & Canada)
- No Admission Charge