Students should be able to to:

• Understand the motivations for the need to integrate learning, reasoning and optimisation, and the role of prior knowledge and knowledge representation.
• Understand integrated representations for trustworthy AI.
• Understand different paradigms that integrate different representations. In particular:
  • Statistical relational AI: the integration of logic and probability/fuzziness for both reasoning and learning.
  • Neurosymbolic AI: integrating logic with neural networks to enable perception and reasoning.
  • Knowledge graphs, ontologies, graph neural networks and embedding.
  • Constraint satisfaction and optimisation techniques: integrating solvers and learners for better performance and for learning CSP models.
• Apply the above methods in perception, spatial reasoning, natural language processing, vision, and other societal/industrial domains

• Use a wide variety of representations (graphical models, logic, neural networks, knowledge graphs) for both learning and reasoning.
• Have insight into the power and limitations of different types of representations.
• Combine different representations for a particular AI task.
• Use both knowledge and data for a particular AI problem.
• Understand and use the above-mentioned categories of techniques (StarAI, NeSy, CSP, Knowledge graphs, Ontologies (OWL/RDFS).
• Understand the limitations and challenges of the integrated representations and paradigms.
• Understand the trustworthiness of these techniques.

• Work effectively with experts in different learning, reasoning and optimisation paradigms.
• Collaborate with domain experts to identify suitable integrated learning, reasoning and optimisation techniques for trustworthy AI.