Brain Function Through the Lens of Computation (366-0-20)

Instructors

James E Fitzgerald

Meeting Info

Parkes Hall 214: Mon, Wed, Fri 10:00AM - 10:50AM

Overview of class

Understanding brain function is a grand challenge for twenty-first century science that promises revolutionary applications to medicine and artificial intelligence. Mathematical modeling can contribute valuably to this understanding by allowing scientists to formalize experimental findings and reason beyond their intuition.

This course will introduce students to the basic building blocks of neural computation, as well as illustrate how these building blocks combine to generate myriad brain functions. We will begin with an overview of several key principles related to neural network dynamics and neural coding. The bulk of the course will then develop these principles by illustrating how computational neuroscientists have used them to model
specific sensory, motor, and cognitive functions of the brain. For instance, we'll see how neural networks can represent the sensory world, generate movement, or store memories depending on the connections between neurons. We'll also see how these connections can change to enable learning. Over the decades, computational neuroscience has enjoyed a rich dialogue with machine learning and data science, and
lectures interspersed throughout the course will explain the practical importance of computational neuroscience approaches for the modern world.

Computational neuroscience is highly interdisciplinary, and this course is open to students with a wide variety of backgrounds, including those majoring in Neuroscience, Data Science, Physics, Applied Mathematics, and Engineering. Problem sets will use Matlab, and some familiarity with coding is recommended. We recommend that Neuroscience majors complete the core NEUROSCI 202 and 206 courses first.

Registration Requirements

MATH 220-2 or higher or test equivalent. We recommend that Neuroscience majors complete the core NEUROSCI 202 and 206 courses first.

Learning Objectives

* Understand how the fundamental principles of neural computation combine in different ways to generate a diversity of brain functions
* Learn how to write, run, and modify computer code to simulate models and analyze neural data
* Gain exposure to the conceptual and mathematical tools of computational neuroscience
* Solve example interdisciplinary problems at the interface of computational neuroscience, artificial intelligence, and medicine.

Teaching Method

The course will be divided into five Units that will cover a wide range of exciting topics in computational neuroscience.
* Unit 1, Course introduction (1 week): e.g., what is computational neuroscience?, mathematical foundations
* Unit 2, Fundamental elements of neural computation (3 weeks): e.g., neurons, neural coding, synapses, neural networks, learning
* Unit 3, Visual processing (2 weeks): e.g., encoding models, efficient coding, deep neural networks
* Unit 4, Motor control (2 weeks): e.g., central pattern generators, low-dimensional dynamics, recurrent neural networks
* Unit 5, Cognition (2 weeks): e.g., perception, memory, decision making

The stated durations of each Unit are approximate and should only be taken to illustrate the general cadence of the course. The tenth week of the course is set aside as a reading period, and no new content will be presented.

Evaluation Method

Daily Questions (10% of grade): Students will submit responses to a few multiple-choice questions after each class to help them consolidate the key concepts presented during the lecture. The questions will be posted after each class and due before the following lecture. You are encouraged to discuss the questions with other students in the course, but everyone must submit their own response. Do not use generative AI.

Problem Sets (70% of undergraduate grade, 50% of graduate grade): There will be nine weekly problem sets, each of which will count equally to your grade. The problem sets will be assigned and due on Mondays. You are encouraged to collaborate with other students in the course, but everyone must prepare their own solutions. Do not use generative AI.

Research Literature Assessment (20% of graduate grade): Graduate students in the course will read a paper from the computational neuroscience literature, critically assess it, and present it orally to the course instructors (i.e., Dr. Fitzgerald, plus course Teaching Fellows and Teaching Assistants). The specific paper will be selected in consultation with Dr. Fitzgerald and presented at a time during the course that coordinates with the paper's scientific themes.

Final Exam (20% of grade): The final will be a take-home exam. This exam will resemble a problem set. However, you may not collaborate with other students or discuss the problems. Do not use generative AI. The final may include content from any aspect of the course.

Aligned Assessments
The questions posed on the Problem Sets and Final Exam will be designed to support all four learning objectives of the course. The Daily Questions and Research Literature Assessment will not involve computer programming, but they will advance the other three learning objectives.

Class Materials (Required)

I will provide written notes following each lecture, and lecture videos will be posted on Canvas. I will be developing the lectures to be complete and self-contained, and the course will not follow a textbook. Students interested to learn more about a topic covered during the lecture may approach me for additional suggested readings.

Class Materials (Suggested)

None.

Class Attributes

Natural Sciences Foundational Discipline
Natural Sciences Distro Area

Enrollment Requirements

Enrollment Requirements: Students must have completion of MATH 220-2 or test (or transfer) equivalent or higher or graduate standing.

Associated Classes

DIS - Kresge Centennial Hall 2-425: Thurs 3:00PM - 3:50PM