COMP44101 15 Semester 1 Undergraduate Level 6 Michele Caprio Unit coordinator Department of Computer Science 1 FHEQ level (Framework for Higher Education Qualifications) ' Masters/Integrated Masters P4 ' European Credit Transfer & Accumulation System Rating : 7.5 <p>Machine learning is increasingly being used for decision support in data driven applications. A key concept when making decisions based on predictive models is that of uncertainty, e.g., in applications of AI where safety or trustworthiness are required. Uncertainty quantification recognises that exact predictions are often out-of-reach due to theoretical or practical limitations. This module studies different probabilistic machine learning models that incorporate uncertain reasoning and the mathematical concepts and algorithms required to learn such models from data.</p> <p>Machine learning is increasingly being used for decision support in data driven applications. A key concept when making decisions based on predictive models is that of uncertainty, e.g., in applications of AI where safety or trustworthiness are required. Uncertainty quantification recognises that exact predictions are often out-of-reach due to theoretical or practical limitations.</p><p><br/>This module studies different probabilistic machine learning models that incorporate uncertain reasoning and the mathematical concepts and algorithms required to learn such models from data. The unit starts with the concepts behind Uncertainty Quantification as expressed by the language of probability, including multivariate distributions. It later studies different approaches for statistical inference, e.g. Markov Chain Monte Carlo and Variational Inference. After studying the fundamentals of Uncertainty Quantification, the module will cover advanced probabilistic models for several types of data structures such as Probabilistic graphical models, state-space models, Bayesian neural networks and Gaussian processes.</p> <p>The unit aims to introduce the main concepts behind reasoning and learning under uncertainty, including the main machine learning models that are used for uncertainty quantification for different data structures. It will also provide practical experience in applying advanced probabilistic machine learning techniques to real-data problems.</p> <p>1. Describe the fundamental concepts of uncertainty quantification.</p><p>2. Describe the use of uncertainty quantification for active learning, Bayesian optimisation and out-of-distribution learning.</p><p>3. Analyse the differences among statistical inference approaches.</p><p>4. Explain the models and algorithms commonly used in probabilistic graphical models, state space models, Bayesian neural networks and Gaussian processes.</p><p>5. Apply an advanced uncertainty quantification model to a data-driven application using auto-differentiation tools and probabilistic programs.</p><p>6. Critically assess technologies and analyse their suitability for specific application scenarios.</p> Analytical skills Innovation/creativity Project management Problem solving Research Written communication <p>1. Uncertainty quantification</p><p><br/>2. Statistical Inference</p><p><br/>3. Probabilistic graphical models</p><p><br/>4. State-space models</p><p><br/>5. Bayesian neural networks</p><p><br/>6. Gaussian processes</p> <p>- Asynchronous material in the form of video lectures, formative exercises, lecture slides and code examples delivered via the virtual learning environment, lectures and supported tutorials.</p><p>- Weekly lectures consolidating asynchronous materials and providing opportunities for discussion and questions.</p><p>- Supervised lab sessions and tutorials provide support for coursework and formative exercises.</p> 1 Written exam 50% 8 Practical skills assessment 50% <p>Individual and general feedback for the Practical skills assessment.</p><p>Formative assessment in Lectures and Practical sessions.</p> Y <p>- David Poole and Alan Mackworth, Artificial Intelligence: Foundations of Computational Agents, Third Edition, Cambridge University Press, 2023</p><p>- Kevin Murphy, Probabilistic Machine Learning: Advanced Topics, First edition, The MIT Press, 2023.</p><p>- Osvaldo Martin, Ravin Kumar, Junpeng Lao, Bayesian Modeling and Computation in Python, CRC Press, 2022.</p><p>- Christopher Bishop, Pattern Recognition and Machine Learning, Springer, 2007.</p> Assessment written exam 2 Lectures 15 Practical classes & workshops 10 Tutorials 5 150