Machine Learning and Optimisation
Occurrence | Teaching cycle |
---|---|
A | Spring Term 2022-23 to Summer Term 2022-23 |
This module advances the IS stream by introducing the basics of machine learning, purely from an optimisation perspective. The range of topics covers linear regression (picking up where Data 1 left off) to decision trees and a simple neural network (leading into advanced machine learning in the third year). Understanding ML requires knowledge of some mathematical concepts that build upon A-level standard mathematics, specifically: Linear Algebra and Continuous Optimisation. This will be taught in-place. Students will see motivating real world problems, the ML techniques required to solve them, the underlying mathematics needed for the technique and their practical implementation. Practicals will be taught using a python-based modern machine learning library such as TensorFlow or PyTorch and so students will gain experience with the declarative programming paradigm (building on the Software stream).
I201 |
Define the machine learning paradigm and distinguish it from the wider field of AI |
I202 |
Distinguish supervised from unsupervised learning, parametric from nonparametric learning and classification from regression. Be able to express a classification problem as a regression problem. |
I203 |
Compute partial derivatives and understand the concept of the gradient as a generalisation of the derivative |
I204 |
Express, manipulate and solve systems of linear equations using linear algebra |
I205 |
Apply linear regression and logistic regression with one variable or multiple variables |
I206 |
Optimise multivariate functions using gradient descent |
I207 |
Explain the concept of overfitting and how regularisation can be used to prevent it |
I208 |
Construct a basic neural network and learn its weights via optimisation using the backpropagation algorithm |
I209 |
Identify clusters in data by applying the unsupervised k-means algorithm |
I210 |
Use K-nearest neighbours and decision tree learning for nonparametric classification |
I211 |
Use a modern machine learning library to implement basic machine learning methods as a computational graph and apply them to real datasets |
I212 |
Appreciate the ethical and privacy implications of machine learning such as relying on black box systems without interpretable parameters, learning from biased or crawled publicly available data |
Task | Length | % of module mark |
---|---|---|
Groupwork INT2 Open Assessment |
N/A | 50 |
Online Exam -less than 24hrs (Centrally scheduled) Intelligent Systems 2 (INT2) Exam |
3 hours | 50 |
None
Task | Length | % of module mark |
---|---|---|
Groupwork INT2 Open Assessment |
N/A | 50 |
Online Exam -less than 24hrs (Centrally scheduled) Intelligent Systems 2 (INT2) Exam |
3 hours | 50 |
Feedback is provided through work in practical sessions, and after the final assessment as per normal University guidelines
Solomon, Justin. Numerical Algorithms. AK Peters/CRC Press, 2015