Overview

This module is divided into two main parts: machine learning fundamentals and modern practices. In the first part, we will cover the basics, including regression, classification, structural analysis, and density estimation. The second part will focus on modern topics, particularly Neural Networks and Deep Learning, which have become the preferred methods for solving various machine learning problems.

Indicative content includes:

• Explain and apply the fundamental notions and principles of machine learning.
• Detail and apply various regression and classification models.
• Detail and apply density estimation and clustering algorithms to data sets.
• Explain methodologies such as overfitting, underfitting, error estimation, ROC curves, transformation, and Ensembles.
• Introduction to neural network models and their training procedures.
• Explain tensor data processing
• Explain linear regression with automatic differentiation,
• Explain single-layer neural networks, and multilayer perceptrons.
• Explain convolutional neural networks.

Assessment Strategy

-threshold -Equivalent to 50%. Uses key areas of theory or knowledge to meet the Learning Outcomes of the module. Is able to formulate an appropriate solution to accurately solve tasks and questions. Can identify individual aspects, but lacks an awareness of links between them and the wider contexts. Outputs can be understood, but lack structure and/or coherence. -good -Equivalent to the range 60%-69%.Is able to analyse a task or problem to decide which aspects of theory and knowledge to apply. Solutions are of a workable quality, demonstrating understanding of underlying principles. Major themes can be linked appropriately but may not be able to extend this to individual aspects. Outputs are readily understood, with an appropriate structure but may lack sophistication. -excellent -Equivalent to the range 70%+. Assemble critically evaluated, relevant areas of knowledge and theory to construct professional-level solutions to tasks and questions presented. Is able to cross-link themes and aspects to draw considered conclusions. Presents outputs in a cohesive, accurate, and efficient manner.

Learning Outcomes

• Apply the machine learning taxonomy to formulate meaningful questions, identify appropriate techniques to address them, and compare and contrast them.

• Describe the theoretical underpinnings of Neural Networks, gradient-based optimisation and automatic differentiation as a tool for modern AI.

• Design a Machine Learning pipeline using PyTorch, one of the most popular Deep Learning frameworks in both Academia and Industry.

• Develop and compare structural analysis (clustering) and density estimation (unsupervised learning).

• Develop and compare various regression and basic classification models (supervised learning).

• Develop the methodology needed to build and evaluate machine learning solutions.

• Discuss the basic notions and principles of machine learning.

• Implement several widely-used Deep Learning architectures.