Prof Milan Vojnovic

This course is available on the BSc in Actuarial Science and BSc in Mathematics, Statistics and Business. This course is available with permission as an outside option to students on other programmes where regulations permit and to General Course students.

ST102 Elementary Statistical Theory

A computer programming course, e.g. ST1aa Programming for Data Science

The objective of this course is to introduce students to basic principles of artificial intelligence systems. By aritificial intelligence, we refer to machines (or computers) that mimic cognitive functions that humans associate with the human mind, such as learning and problem solving. The course will take a practical approach, explaining the main principles and methods used in the design of artificial intelligence systems.

The course will cover different types of learning, including supervised, semi-supervised, unsupervised and reinforcement learning; neural network architectures used for solving tasks such as image classification, speech recognition, and neural machine translation; reinforcement learning problem formulation using the framework of Markov Decision Processes, elementary solution methods based on tabular approaches (maintaining estimates of state values or state-action values) and those based on using function approximation for value functions using deep learning models. Students will gain practical knowledge by solving exercises using modern, commonly used software libraries such as TensorFlow and OpenAI Gym.

20 hours of lectures and 15 hours of seminars in the LT.

Students are required to use a Python programming environment, e.g. by installing Anaconda / Jupyter notebooks on their laptops, or using Google Colab. 

Students not having a laptop of their own, which they can use for the purpose of the course, will be offered to use personal computers available in seminar rooms. 

Students will be expected to produce 10 exercises in the LT.

In each teaching week, students will be given exercises that will contain problem solving questions to check their comprehension of the theoretical concepts and programming exercises for gaining practical hands-on problem solving skills. 

1. I. Goodfellow, Y. Bengio and A. Courville, Deep Learning, MIT Press, 2017, http://www.deeplearningbook.org
2. R. Sutton and A. C. Barto, Reinforcement Learning: An Introduction, 2nd Edition, MIT Press, 2018 
3. S. Russell and P. Norvig, Artificial Intelligence: A Modern Approach, 3rd Edition, Pearson, 2016
4. TensorFlow, An Open Source Software Library for Machine Intelligence, http://www.tensorflow.org
5. OpenAI Gym documentation, http://gym.openai.com/docs/

1. T. Hastie, R. Tibshirani, and R. Friedman. The Elements of Statistical Learning: Data Mining, Inference, and
   
   Prediction. 2nd Edition. Springer. 2009, available online at https://web.stanford.edu/~hastie/Papers/ESLII.pdf
2. M. Wooldridge, An Introduction to MultiAgent Systems, 2nd Edition, Wiley, 2009
3. L. Deng and D. Yu, Deep Learning: Methods and Applications, Now Publishers Inc, 2014
4. F. Chollet, Deep Learning with Python, Manning, 2018
5. A. Geron, Hands-on Machine Learning with Scikit-Learn and Tensorflow, O'Reilly, 2017

Exam (50%, duration: 2 hours) in the summer exam period.
Coursework (50%) in the LT.

Students are required to hand in the solutions to 2 sets of exercises (each accounting for 10% of the final assesment) and 1 set of problem solving tasks using Python (which accounts for 30% of the final assessment). 

The course will cover several theoretical concepts, such as optimisation methods for learning deep neural networks, and the framework of Markov Decision Processes (MDP) for formulation of reinforcement learning problems. The knowledge of these theoretical concepts is best assesed by an exam.