Skip to main content

Advanced Machine Learning:  2013-2014

Lecturer

Degrees

Schedule C1Computer Science

Schedule C1Mathematics and Computer Science

Schedule CMSc in Advanced Computer Science

Term

Overview

Machine learning techniques enable us to automatically extract features from data so as to solve predictive tasks, such as forecasting, missing data imputation, anomaly detection, classification, ranking, control and decision making. Learning systems adapt so that they can solve new tasks, similar to previously encountered tasks, more efficiently.

This course covers the main machine learning methods for dealing with problems where human expertise is absent (e.g., robots solving problems in other planets), where humans are unable to explain their expertise (e.g., object recognition), where solutions depend on a context (e.g., personalization, biometrics and tracking), and where the problem size is too vast for the limited reasoning capabilities of humans (e.g., web ranking and automatic tuning of large systems). The course also introduces hard topics such as learning representations, theories and causal relations. The practicals will concern the application of machine learning to a range of real-world problems and datasets.

Learning outcomes

On completion of the course students will be expected to:

  • Have a good understanding of the two numerical approaches to learning (optimization and integration) and how they relate to maximum likelihood and the Bayesian approach.
  • Have an understanding of how to choose a probabilistic model to describe a particular type of data.
  • Know how to evaluate a learned model in practice.
  • Understand the role of machine learning in massive scale automation.
  • Have a good understanding of the problems that arise when dealing with very small and very big data sets, and how to solve them.
  • Understand the mathematics necessary for constructing novel machine learning solutions.
  • Be able to design and implement various machine learning algorithms in a range of real-world applications.

Prerequisites

Machine Learning is a mathematical discipline, and students will benefit from a good background in probability theory, linear algebra and continuous mathematics, such as is covered in the relevant Prelims courses.  Students will be expected to have taken the Machine Learning course

Synopsis

  • 1. Introduction (1 lecture)
  • 2. The optimisation approach to linear regression (2 lectures)
    • a. Maximum likelihood estimates (MLE)
    •  b. Ridge regression
    • c. Lasso and variable selection
    • d. Sub-gradients and the shooting method
    • e. Cross-validation and the bootstrap
    • f. Matrix factorization and collaborative filtering
  • 3. Optimization (2 lectures)
    • a. Deterministic and stochastic gradient descent
    • b. Newton’s method and variants
    • c. Nesterov’s method
    • d. Duality
    • e. Proximal methods and distributed optimization
  • 4. The Bayesian approach to linear regression (1 lecture)
    • a. The multivariate Gaussian
    • b. Conjugate analysis
  • 5. Generalised linear models and the exponential family (1 lecture)
    • a. Link functions (logit, probit)
    • b. MLE and MAP estimates
    • c. Multi-task learning
    • d. Learning to rank
  • 6. Approximate inference (2 lectures)
    • a. Variational Bayes
    • b. Monte Carlo
  • 7. Deep learning (3 lectures)
    • a. Neural networks and backpropagation
    • b. Autoencoders for unsupervised and transfer learning
    • c. Boltzmann machines and score matching
    • d. Randomization
    • e. Convolutional pooling architectures
  • 8. Gaussian processes (1 lecture)
  • 9. Bayesian optimization, online learning and bandits (2 lectures)
    • a. Maximum expected utility and acquisition functions
    • b. Regret and online learning
    • c. Automatic algorithm configuration
    • d. Thompson sampling
    • e. Markov decision processes and reinforcement learning
  • 10. Randomized methods for learning (2 lecture)
    • a. Basic concentration inequalities
    • b. Johnson-Lindenstrauss theorem
    • c. Random projections, hashing and sketches
  • 11. Probabilistic graphical models (2 lectures)
    • a. DAGs, HMMs and Kalman filters
    • b. Random fields and factor graphs
    • c. Conditional random fields
  • 12. Ensemble methods (2 lectures)
    • a. random forests b. boosting
  • 13. Learning causal models (1 lecture)
    • a. Counterfactuals
    • b. Instrumental variables

Syllabus

Mathematics of machine learning. Overview of supervised, unsupervised, multi-task, transfer, active and reinforcement learning techniques.

Reading list

Primary Text 

  • Kevin P. Murphy. Machine Learning: A Probabilistic Perspective, MIT Press 2012.

 

Secondary Texts

  • Christopher M. Bishop. Pattern Recognition and Machine Learning, Springer 2007.
  • T. Hastie, R. Tibshirani, and J. Friedman. The Elements of Statistical Learning. Springer 2011.

Feedback

Students are formally asked for feedback at the end of the course. Students can also submit feedback at any point here. Feedback received here will go to the Head of Academic Administration, and will be dealt with confidentially when being passed on further. All feedback is welcome.

Taking our courses

This form is not to be used by students studying for a degree in the Department of Computer Science, or for Visiting Students who are registered for Computer Science courses

Other matriculated University of Oxford students who are interested in taking this, or other, courses in the Department of Computer Science, must complete this online form by 17.00 on Friday of 0th week of term in which the course is taught. Late requests, and requests sent by email, will not be considered. All requests must be approved by the relevant Computer Science departmental committee and can only be submitted using this form.