《Machine Learning and its Applications in Chemistry》course for undergraduates and graduates

Credit: 2;  Class hours: 32.  Lecturer: Prof. Zhirong Liu

Background:

      Machine learning is an interdiscipline which studies how computer program learns from data in order to acquire new knowledge and to improve the performance. Machine learning has a long history in chemistry, e.g., DENDRAL was an expert system developed in the 1960s which can identify unknown organic molecules by analyzing their mass spectra and using knowledge of chemistry. In the past decade, with the rapid development of deep learning, revolutionary breakthroughs were achieved in the fields of machine learning and artificial intelligence. An eye-catching example is the success of AlphaGo, an AI computer program to beat the human world champion in the Go game. Nowadays, machine learning has permeated into various fields, including chemistry: it has been successfully applied in organic syntheses (M. H. S. Segler et al., Planning chemical syntheses with deep neural networks and symbolic AI, Nature 555, 604 (2018)) and protein-structure prediction (AlphaFold was ranked the first in the 13th Critical Assessment of Structure Prediction (CASP) in 2018, a biannual competition aimed at predicting the 3D structure of proteins). In the course, we will give an  introduction to the vast field of machine learning (basic concepts, methods and workflow) and survey how machine learning can be used in chemistry. The general aim is to convey the exciting potential of machine learning, which you may meet or employ in your own work later.

Aims:

      Driven by the development of deep learning, the field of machine learning and artificial intelligence has made rapid development in recent years, which has begun to have an impact on many disciplines, including chemistry. In view of the challenges and opportunities that machine learning may bring to chemistry, this course will introduce the basic knowledge of machine learning, and analyze the typical application examples of machine learning in the frontier of chemistry. Through the course, we hope that students can achieve the following goals: (1) master the basic knowledge of machine learning; (2) be able to read, analyze and evaluate the machine learning literature in the field of chemistry; (3) be able to use machine learning to solve simple problems.

References:

1、Christopher M. Bishop, Pattern Recognition and Machine Learning. (Springer, 2006).

2、Richard S. Sutton and Andrew G. Barto, Reinforcement Learning: An Introduction. 2nd edition. (MIT Press, 2018).

3、Ian Goodfellow, Yoshua Bengio, and Aaron Courville, Deep Learning. (MIT Press, 2016).

4、Peter Harrington, Machine learning in action (Manning Publications, 2012).

Contents:

* Introduction to machine learning

* Linear regression: the simplest machine learning model

* Linear method of classification: logistic regression

* Artificial neural network

* Model assessment

* Kernel method:  support vector machine

* Graph model: naive Bayesian classifier; Bayesian network; Markov random field; Boltzmann machine

* Clustering: K-means method; Gaussian mixture model

* Dimension reduction:  PCA

* Ensemble learning: boosting; decision tree; random forest

* Reinforcement learning 1: multi-arm slot machine

* Reinforcement learning 2: finite Markov decision process

* Reinforcement learning 3: dynamic programming, Monte Carlo simulation and temporal-difference method

* Reinforcement learning 4: AlphaGo

* Deep learning

* Chemical examples: organic synthesis routes, drug design, force fields