While AI has made rapid progress, there still remains a large gap between machine intelligence and natural intelligence. In this course, you will learn about recent developments in the fields of machine learning and neuroscience that provide insights into the principles and mechanisms that govern natural intelligence. Ultimately, this leads to a better understanding of neural information processing in biological systems and paves the way for the development of more capable machine intelligence. After completion of the course, the student is able to:
- Understand the principles that shape neural information processing systems
- Describe relevant models and theories in machine learning and neuroscience
- Implement computational models and algorithms for learning, inference and control in neural systems
- Train artificial agents to maximise reward in complex environments
|
|
|
We consider natural agents as adaptive systems that strive to maximise reward in complex stochastic environments. We formalise the interaction between agents and their environments in terms of partially-observable Markov decision processes. Next, we investigate how neural information processing systems are shaped by neuron models, objective functions, learning algorithms and network architectures. Finally, we consider how natural agents can learn to solve tasks using recent developments in e.g. reinforcement learning and self-supervised learning. You will learn to implement several models and algorithms and we will discuss recent papers in the field.
|
|
|
|
This course requires experience with the Python programming language, prior experience with neural networks and basic mathematical skills (calculus, linear algebra, probability theory).
|
|
- Practical assignment (50%)
- Written exam (50%)
There is only a resit for the written exam.
|
|
|