"GTIIT Soft Matter and Biological Physics Seminar" -- "GTIIT BioSoft seminar" in brief, is initiated in Nov. 2018 to build up the connections of GTIIT with other scientists in the interdisciplinery research field: Soft matter and biological physics. We will invite some of the best physicists, material scientists and applied mathematicians from inside/outside China to give talks or series of lectures. Although the talks may be too advanced for first 2 years’ undergraduate students, yet they are completely welcome to gain a helpful/healthy feeling about the frontiers of modern interdisciplinary scientific researches.

We are pleased to announce that the Workshop on Machine Learning for Soft Material Science will be held at 16:00 on May 11th, 13th, and 15th, 2020. The workshop will bring together researchers in and out of China to give introduction lectures to machine learning, share the latest advances and research prospects in the development of machine learning approaches in physical science and to explore their potential applications in the study of soft matter physics. The lectures will be open to all GT students, but pre-registration will be needed due to the limit of online courses.

机器学习是描述如何从大数据中学习和做出预测的领域。近年来，机器学习是科学研究和应用中最令人兴奋，也是发展最迅速的领域之一，它的概念和方法正在被应用到越来越多的科学和工程技术研究中。本项主题研讨会旨在探讨机器学习内在的基础数学理论，为进一步提高其计算效率提供新思路；讨论机器学习具体在软材料科学、复杂流体问题、以及生命科学的数据分析计算中的应用潜力和所面临的挑战。组织人拟邀请5位在机器学习领域有突出贡献的数学专业、计算机专业及相关材料科学专业方面的学者进行研讨，推动多学科交叉融合和密切合作交流。

A First Introduction to Machine Learning

Lecturer: Xiang ZHOU, Associate Professor, City University of Hong Kong

This 3-hour tutorial is a mathematical introduction to (classic) statistical machine learning for the audience in mathematics, science and engineering majors but without statistical or machine learning background. The pre-requisits are basic probability and linear algebra. After the course , the audience are expected to understand the basic concepts and problems of statistical learning and machine learning in a rigorous and unified approach. Roughly speaking, the main idea of this tutorial is more like ‘modelling’ process in physics instead of showing a vast number of numerical methods, although the latter is undeniably important in practice. However, due to the limited time, any practical algorithms unfortunately have to be skipped. It is preferred for the audience to continue to study (more detailed) numerical algorithms with practical hands-on experiences for some applied problem of their own interest. The tutorial is split into three sessions:

Session one: learning as approximation; generalization error; bias-variande trade off, cross-validation. This is the most important part since I try to explain the big and unique picture of machine learning.

Session two: linear and nonlinear regression (Generalized Linear Models), including variable selection (ridege, lasso).

Session three: classification problem; naive Bayesian; logistic regression; loss functions in classification.

Warning: This tutorial is not for those practitioners who are looking for a quick practical solution to applications.

Course prerequisite:

Linear Algebra, Probability, Python basics, Statics (optional), Optimization (optional)

Student quota for pre-registration (first come, first served):  30

(The first 30 students who replied the email to UG office will be registered automatically and ONLY successful registration will be notified and given a Zoom meeting ID).

Organizer and secretary:

Xinpeng XU                   Guangdong Technion – Israel Institute of Technology