ÃÀ¼§¸ó

This module aims to equip you with practical knowledge in statistics required for assessment and quantification of uncertainties in real life scenarios of data analysis. Module presents practically important algorithms of statistical learning for both prediction and decision making purposes and provides the opportunities for your experimental evaluation during the lab sessions. Tools for evaluation of learning algorithms’ performance are also considered and implemented to practical examples.

• Introduction to statistical learning,
• Statistics fundamentals: probability, random variables, descriptive statistics and stochastic processes,
• Statistical inference: estimation and testing, evaluation metrics,
• Bayesian methods: Naïve Bayes and Bayesian Networks,
• Markov processes and chains, Kalman estimators,
• Statistical modelling and decision making: regression, mixture models and classification approaches,
• Case study: application of statistical learning for aerospace sector problem.

On successful completion of this module you should be able to:

• Relate statistical techniques for uncertainty quantification to real life problems,
• Differentiate experimental data according to the underlying models of stochastic processes,
• Propose statistical learning methods suitable for particular problem,
• Assess the outcomes of the statistical learning.

The module aims at giving you a solid background introduction to optimization and decision theory. The use of modern optimisation methods, especially for applications in artificial intelligence. More than the traditional linear techniques, non-linear techniques will also be addressed, including multi-criteria methods. 

• Introduction to Optimisation and Decision Theory  
• Optimisation models and methods  
• Integer and Mixed-integer programming 
• Linear and non-linear programming (including intro to Meta-heuristics) 
• Decision analysis  
• Multiple-criteria decision analysis.

On successful completion of this module you should be able to:

1. Formulate decision problems based on optimisation scenarios, identifying the different variables. 
2. Differentiate and apply different optimisation models and methods. 
3. Evaluate optimisation and decision problem criteria. 
4. Design decision analysis problems and apply/implement different algorithms and solutions.
5. Examine and debate the application of structured approaches to support decision in context of multiple conflicting objectives.

The aim of this module is to introduce the students to advanced deep learning algorithms, architectures, and development platforms, with particular emphasis on the computer vision domain for perception, reasoning, and prediction of real-world objects, suitable for real life problems concerning relevant industrial applications from aerospace, manufacturing or transports.

• Artificial Neural Networks (Shallow models).
• Backpropagation and Training.
• Deep learning architectures.
• Convolutional Neural Networks.
• Recurrent neural networks.
• Deep learning applications in computer vision including object detection, identification, classification, segmentation, tracking, and prediction, etc.
• Generative AI.
• Tensorflow/Pytorch/Matlab practical sessions on Artificial, Convolutional, and Recurrent Neural Networks.

On successful completion of this module you should be able to:

1. Explain fundamental meaning and discuss applicability of deep learning algorithms for industrial applications.
2. Evaluate the commonly used deep learning algorithms, architectures, and platforms in the area of computer vision and describe their applications.
3. Implement deep learning algorithms, estimate their performance in a simulation environment and assess their performance for a realistic case study.
4. Judge AI implementation platforms and create deep learning applications for specific problems in computer vision domain.

The aim of this module is to enable you to think critically about technology, solutions, and gain best practices of intelligent systems issues relating to the cyber-physical systems.

• Cyber-physical systems: Control, sensor and actuators
• Intelligent agent and multi-agent 
• Intelligent robotics  
• Embedded systems 
• Connected system 
• Countermeasures. 

On successful completion of this module you should be able to: 

• Appraise the theoretical and practical aspects for intelligent system in cyber-physical systems approach.
• Distinguish the fundamental aspect of intelligent agent, robotics, multi agent systems. 
• Create working knowledge in dependable control, and embedded systems. 
• Assess key issues of connected system within the physical world.
• Analyse different approaches of cyber-physical system with consideration of countermeasures.

• Dr Ivan Petrunin

This module will introduce students to data analytics, overview challenges and solutions in this area, present approaches to predictive and descriptive data mining and explain unsupervised learning techniques suitable for new information discovery. Visualisation tools and performance metrics are also considered within the module. You will benefit from knowledge of basic concepts of statistics for performance assessment and evaluation.

• Introduction to Data Analytics,
• Data exploration and pre-processing,
• Predictive analytics: regression and classification methods,
• Clustering and dimensionality reduction,
• Graph analysis and visualisation,
• Software and tools for data analytics,
• Case study: application of data analytics techniques and visualisation tools for knowledge discovery problem.

 

On successful completion of this module you should be able to:

• Distinguish stages of the data analytics workflow,
• Categorize data analysis and visualisation techniques with respect to data analytics stages,
• Plan data analytics workflow based on the available data and formulated requirements,
• Set up algorithms for discovery of new information from the large data sets,
• Evaluate performance of the algorithms and quality of the data analysis outcomes.

The aim of this module is to provide the you with some of the most used deep learning techniques applied in the decision-making process of autonomous systems. The module presents practically important algorithms used for real-world human-machine, machine-environment, and human-machine-environment applications in transportation, manufacturing and aerospace.

• • Recurrent Neural Networks.
• • Natural Language Processing.
• • Large Language Models.
• • Introduction to Reinforcement Learning.
• • Deep Reinforcement Learning.
• • Advanced Reinforcement Learning.
• • Graph Neural Networks.
• • Tensorflow/Pytorch practical sessions on Natural Language Processing, Large Language Models, Reinforcement Learning and Deep Reinforcement Learning.

On successful completion of this module you should be able to:

1. 1. Identify the main elements of deep learning algorithms applied in autonomous decision-making applications.
2. 2. Assess the commonly used deep learning methods and describe their advantages, disadvantages and suitability in different applications.
3. 3. Propose deep learning methods and/or create deep learning methodologies suitable for specific real-world problems.
4. 4. Evaluate the performance of deep learning methods in a simulated environment using appropriate metrics.

This module aims at providing you with the foundations of propositional logic as fundamental formal language to represent relevant knowledge. It introduces the main basic techniques and methodologies for logical reasoning to design automated reasoners and it provides the basics of Formal Modelling for engineering systems, a well-established industrial tool to check systems designs via their digital twins.

It also introduces Causal Logic linking it to the founding principles of Machine Learning.

Introduction to logic representation and reasoning

Logic Agents

Propositional Logic

Inference Algorithms

First Order Logic

Exercises and case studies

Formal Models

Formal Verifications

State machines

Exercises and case studies

Causal inference and machine learning

On successful completion of this module you should be able to:

1. 1. Define the relevant Knowledge Base (KB) to represent a given problem, and design it using an appropriate knowledge representation language.
2. 2. Apply inference to code autonomous reasoning algorithms.
3. 3. Analyse engineering systems and their processes to model them as state machines.
4. 4. Design relevant autonomous model checkers for modelled systems.
5. 5. Apply causal inference techniques to identify and work with the causal relationships between system parameters and how these techniques relate to machine learning.

Technology typically advances much faster than society develops an understanding of its potential consequences, and before appropriate regulations and standards are developed to prevent negative impacts. This module aims to provide you with a range of knowledge from different ethical, regulatory and social perspectives that will prepare them for the advancing landscape of AI. It will address fundamental theoretical and practical ethical issues (e.g. data management and security, personal control and consent, biases, etc.) and explore them in relation to current standards and regulatory frameworks. In addition, the module will cover a range of key social science topics and techniques to equip you with an appreciation of how to optimise AI integration and the ability to apply practical methods to do so. 

The sessions in this module will focus on three inter-related strands: ethical, regulatory, and social aspects of AI.

• Ethical aspects: exploring and debating ethical principles for AI, from fundamental definitions and the development of moral norms and expectations in society to more specific appraisal of how ethics will shape, and be shaped by, the developing landscape of intelligent systems. Key concepts will be appraised and related to context via classroom activities and discussions to predict possible and probable challenges in dealing with ethical behaviours in the design and application of AI and the potential human / social impacts.
• Regulatory aspects: examining and evaluating the current state of regulations for AI, how formal standards and guidance specifications are developed, and how they are harmonised and linked to legal frameworks. Existing standards / regulatory documents will be reviewed (subject to availability and access permissions) to facilitate group and individual appraisals of content and usefulness in practice, and to predict potential gaps with respect to the evolving landscape of AI in society today and in the future. 
• Social aspects: identifying and assessing the various current and future social implications of AI, with particular focus on the bi-directional impacts between systems and the surrounding cultural and social environment. Psychological processes and human factors that influence attitudes and behaviours towards intelligent systems and their applications in society will be studied, along with practical human / social data collection, measurement and analysis techniques.