UROP Openings

Have a UROP opening you would like to submit?

Please fill out the form.

Submit your UROP opening

Structure learning using 3d spatial environments


Term:

Summer

Department:

9: Brain and Cognitive Sciences

Faculty Supervisor:

Ila Fiete

Faculty email:

fiete@mit.edu

Apply by:

June 1, 2020

Contact:

susharma@mit.edu

Project Description

Grid cells[1], a type of nerve cell in the brain, form a part of its navigational GPS. They tessellate space in a triangular lattice and fire at the vertices of this lattice. They have a rich phenomenology [2] and have a huge coding capacity for representing space [3] due to their modularity, i.e., they exist in discrete clusters in the brain. The spatial relationships between firing fields of grid cells remain constant across environments, suggesting that these cells are part of a general metric representation of space. An interesting phenomenon exhibited by grid cells is the phenomenon of remapping. For instance, in a hairpin maze, grid representations break down into discontinuous segments with fragmentation occurring abruptly near the point of turning [4]. This implies that grid cells form discrete submaps for each part of the hairpin maze environment. This project will involve modelling 3D environments in Unity (video game engine) to probe questions about the metric representations and structure learning in Humans through behavioural experiments. The goal would be to relate behavioural variables to grid cell properties. Interested students will learn how to build interactive 3D environments, run a behavioural experiment on Amazon mechanical turk, and possibly explore aspects of planning in realistic 3D environments. Ideal for people looking to build something cool this summer while learning about human cognition and it’s underlying mechanisms. [1] http://www.scholarpedia.org/article/Grid_cells [2] Stensola et al, Nature (2012), The entorhinal grid map is discretized. [3] Fiete, Burak, Brookings, Journal of Neuroscience (2008), What do grid cells convey about rat position. [4] Derdikman, Dori, et al., Nature neuroscience (2009), Fragmentation of grid cell maps in a multicompartment environment.

Pre-requisites

Proficiency with Python and C#. Prior experience with Unity 3D and Amazon mechanical turk would be a plus. Basic understanding of web development (HTML, CSS, Javascript) would be a plus. This project is suited to students from brain and cognitive sciences, and EECS. Applications from other departments can be considered if a student is interested in the project.