NeuroAI: The Computing Brain

Neural network models of brain and cognition: how the biological brain gives rise to intelligence?

Our current understanding of brain function combines biology, theory and models to explain how neurons and networks underlie behavior. This minor will depart from basics of neuroscience with lectures about neurons, brain anatomy and cognition and build up on that knowledge to teach you how to simulate neurons and neural networks. We will use physical models of the neuron to reproduce neuronal dynamics (“action potentials”) and computational models to reproduce high-level cognitive functions such as pattern recognition, motor behavior, memory and navigation. The student will learn and actively explore the multiple developments of computational neuroscience and artificial intelligence and their relationships with the biological brain.

Some of the guiding questions covered in this minor are:

• How does the neuron work?
• What is computation? What are the differences between biological and in silico computation?
• How does the organization of the nervous system give rise to intelligent behavior?
• How are computational principles implemented in biological substrates?
• How do artificial neural networks help explain brain function?
• How can we build interfaces between brains and computers?

This minor is designed for students with an interest in computational neuroscience and artificial intelligence to dab into modern theories of brain function. The minor emphasizes a mechanistic understanding of brain function and computation by presenting modern neural networks as explanatory devices. We work systematically from essential neurobiology, to neurodynamics, to behavior, onto the fundamentals of artificial neural networks and their application to the problems of intelligence and cognition.

The minor can be taken by students with various backgrounds including medicine, psychology, biomedical sciences, psychology, computer science and philosophy. The problem of explaining brain function is tackled from complementary perspectives, including fundamental biophysics of neurons, neurodynamics, cognition, and from then to models of artificial neuronal networks that emulate brain function. The minor has a strong emphasis on hands on practice via in-class projects and should empower the student with the essential toolbox to understand modern scientific developments of neuronal simulations, neural network models and brain computer interfaces.

This minor is aimed at students with an interest in theoretical aspects of neuroscience (physics, mathematics, computer science) and practical implementations ("coding") of models of neurons and networks. The minor guides the students to explore interfaces between neuroscientific and computational disciplines, with a distinctive emphasis on software implementation of neural network models of brain computation.

Take a look at the video and learn more about this minor.

• Describe the fundamental properties of a biological neuron
• Explain the dynamical process of action potential generation
• Understand how biological neurons are translated into computational models
• Explain different neuronal activity patterns (“Synchrony”) on the basis of biophysical parameters
• Interpret brain signals from different methods of brain measurement (e.g., EEG, fMRI)
• Implement neural networks that perform pattern recognition
• Understand the principles behind modern "deep learning" architectures
• Understand how recurrent neural networks are used to generate spatiotemporal patterns
• Be able to choose between supervised, unsupervised and reward-based learning methods for particular problems
• Explain cognitive functions through via neural network models and analogies (e.g., object recognition / decision making / motor behavior)

This minor is aimed at students from biomedical and technical backgrounds, with an interest in practical implementations ("coding"), to delve into the current state of brain theory.

NB: for students of non technical backgrounds (e.g., medicine, biology, psychology): although the minor is open for students of any backgrounds, it is designed primarily for students with quantitative background and basic coding skills (nanobiology, clinical technology). It is essential that students taking this minor should have keen interest for basic physics, linear algebra, biology and programming.
To help with deciding on whether to take this minor or not, please take the self-evaluation quiz (https://forms.gle/kYubhKQXJWSRJR3CA).
NB: for TU Delft students: if you need to fill your 30 EC we can provide a list with relevant complementary courses.

Teaching methods
The minor includes 1. lectures, 2. programming projects, 3. flipped classrooms. Classes often start with comprehension quizzes and discussion of previous lectures and their overarching context. Students are expected to delve in guided self-study with recommended material (books, articles and video lectures). For the flipped sessions students are expected to procure and share information via online forums. There are two lab visits to the neuroscience department. Computational modeling is mostly conducted in the python language in 'google collaboratory' (students do not need to install python). Students prepare for the final exam via spaced-repetition flash cards (www.brainscape.org).

Teaching material

• Online reading material
• Programming projects (python coding via 'google collaboratory')
• Video-lectures
• Coding tutorials and support sessions
• Flashcards (Brainscape).
• Book chapters from "Principles of Neuroscience"
• Selected scientific articles.

Examination

• Digital Examination – Exam with multiple choice and open questions at the end of week 10.
• Final Project Work and Presentation - In groups of two or individually, students deepen their grasp of the topics in a chosen project. It is assumed that students have heterogeneous backgrounds. In the project presentations we expect the students to be able to demonstrate their learning arc. They must be able to show how the projects involve the new knowledge gained in the course.
• Each student must present at the each of the 3 flipped sessions (Electrophysiology / Synapses / Brain Computer Interfaces). Each of these sessions accounts for 10% of the grade (pass/fail)

Composition final grade
Important to note:

• To pass the minor students must achieve a minimum grade of at least 5.5 in each of the components.
• Digital Examination (testvision) – 30% (Multiple choice and open questions)
• Final Project (Evaluated via Report, peer + lecturers) – 30%
• Flipped sessions (3-5 sessions, each pass/fail) – 30%
• Engaged participation grade - 10% (with feedback from lecturers)