This course is an introduction to the core concepts in machine learning with an emphasis on applications in physical sciences, including quantum physics, statistical mechanics, and quantum information processing. Lectures will cover the basic theory underlying several modern areas of machine learning and have practical components wherein applied techniques in machine learning are discussed. Students are expected to be comfortable with programming in Python 3.

 

Develop a basic understanding of learning algorithms and their areas of application;

Get acquainted with practical ingredients of machine learning projects;

Develop required skillset for performing a team-wide software project;

Get involved in reading and understanding machine learning literature;

Develop the skillsets required for reproducing the results reported in machine learning literature.

• Introduction to learning and learning paradigms
• Fitting and overfitting
• Linear regression
• Logistic and softmax regression
• Regularization and generalization power
• Deep learning:
• Classification and recognition tasks
• Feedforward and backpropagation
• Stochastic gradient descent
• Convolutional neural networks
• Recurrent neural networks
• Statistical learning:
• Support vector machines
• Structured prediction
• Generative models:
• Variational auto-encoders
• Generative adversarial networks
• Energy-based models

Additional topics, time allowing:

• Representational power
• Bayesian optimization
• Unsupervised learning
• Reinforcement learning

 

Programming resources

 

Title / Name	Notes / Comments	Required
A computing platform for programming	E.g. a laptop.	Yes
A Github account	The free account is sufficient (http://www.github.com/)	Yes

 

More resources

 

Title / Name	Notes / Comments	Required
Deep Learning by I Goodfellow, Y Bengio, and A Courville		No
Stanford CS 229 notes	http://cs229.stanford.edu/syllabus.html	No
UBC CPSC 540 notes	https://www.cs.ubc.ca/~schmidtm/Courses/540-W19/	No
Applying machine learning to physics	https://physicsml.github.io/pages/papers.html	No

All assignments and the final project are done through Github. The student is required to make an account on Github and provide the account ID to the instructor/TA of the course.

Deep Learning (Goodfellow, Bengio, Courville) is perhaps as close as a textbook can get to overlap the topics of this course. Yet many topics discussed here will be skipped and several topics of our focus are out of the content of this book. Similar to other ML textbooks, this book comes in hundreds of pages, so perhaps an introductory short read to general topics in ML are The Hundred-Page Machine Learning Book (by Andriy Burkov) and Neural Networks and Deep Learning (by Michael Nielsen). None of these books are required material for the course.

For some more advanced topics refer to Stanford CS 229 and UBC CPSC 540. We will touch some of the material discussed in these courses as they are relevant to applications in physics.