Perspective module from Wang et al 2023

Motivation

We train models to predict the visual cortex responses for mice (a model takes an image or a video as input and predicts the neuronal response). We also use behavioural variables such as mouse running speed or pupil position and dilation as they are known to be the modulators for the neuronal responses. Currently, the core takes the stimuli and some behaviour as input and pupil position is an input to the readout, which follows the core. However, it would make sense to process the pupil position before the core as it essentially says where the animal is looking.

Project

The minimum goal of the project is to adopt the perspective module from Wang et all, 2023 into the neuropredictors framework. The further idea could be to adjust the module to make it sharable between sessions and animals.

Thesis

Within the context of this project, several research questions could be explored: - Does the perspective network improve the CNNs performance for our core architecture? Why? - Does it work for both images and videos? - Can we use same pespective network for several animals? Currently, in the foundational model it is per animal. Are there any modifications needed for this?

Contact

To apply please email Polina Turishcheva stating your interest in this project and detailing your relevant skills. A part of this project could be also a lab rotation.