Events
FDS Colloquium
Towards Interpretable, Robust Trustworthy Machine Learning for Diverse Applications in Science and Engineering
 |
Speaker: Guang Lin (Purdue) Associate Dean for Research and Innovation and the Director of Data Science Consulting Service Purdue University Wednesday, December 11, 2024 11:30AM - 1:00PM Lunch at 11:30am in 1307
Talk at 12:00pm in 1327 |
Add To: Google Calendar | Outlook | iCal File
Abstract: This talk aims to create new technologies that can be translated into interpretable, robust trustworthy AI systems that can be deployed for real-time complex dynamical system prediction, and applications to improve the stability and efficiency of complex dynamical systems. In the first part of the talk, I will employ Covid-19 pandemic prediction and personalized prediction in Alzheimer’s disease to illustrate how to build an interpretable, trustworthy data-driven model. In the second part of the talk, I will introduce scalable algorithms for Bayesian deep learning via Repica exchange stochastic gradient Monte Carlo. Replica exchange Monte Carlo (reMC), also known as parallel tempering, is an important technique for accelerating the convergence of the conventional Markov Chain Monte Carlo (MCMC) algorithms. However, such a method requires the evaluation of the energy function based on the full dataset and is not scalable to big data. The naïve implementation of reMC in mini-batch settings introduces large biases, which cannot be directly extended to the stochastic gradient MCMC (SGMCMC), the standard sampling method for simulating from deep neural networks (DNNs). In this talk, I will present an adaptive replica exchange SGMCMC (reSGMCMC) to automatically correct the bias and study the corresponding properties. The analysis implies an acceleration-accuracy trade-off in the numerical discretization of a Markov jump process in a stochastic environment. Empirically, we test the algorithm through extensive experiments on various setups and obtain the state-of-the-art results on CIFAR10, and CIFAR100 in both supervised learning and semi-supervised learning tasks.
Speaker bio: Prof. Guang Lin is the Associate Dean for Research and Innovation and the Director of Data Science Consulting Service that performs cutting-edge research on data science and provides hands-on consulting support for data analysis and business analytics. He is also the Chair of the Initiative for Data Science and Engineering Applications at the College of Engineering. Guang Lin is also a Full Professor in the School of Mechanical Engineering and Department of Mathematics at Purdue University.
Lin received his Ph.D. from Brown University in 2007 and worked as a Research Scientist at DOE Pacific Northwest National Laboratory before joining Purdue in 2014. Prof. Lin has received various awards, such as the NSF CAREER Award, Mid-Career Sigma Xi Award, University Faculty Scholar, College of Science Research Award, Mathematical Biosciences Institute Early Career Award, and Ronald L. Brodzinski Award for Early Career Exception Achievement.
Submit an Event
Interested in creating your own event, or have an event to share? Please fill the form if you’d like to send us an event you’d like to have added to the calendar.