Yue Xu

Project #1 (ongoing)

Develop algorithms to explain traits and states using human behavioral data

Subtopic #1

Traits and states are latent variables that serve to explain regularities in human and animal behavior. Traits are temporally stable, like personality traits, and states are temporally phasic, like momentary emotions. My goal is to develop algorithms to extract these latent variables from human behavior data such as the dataset collected by the COVID-Dynamic longitudinal study. (image source)

Subtopic #2

Apply algorithms on data to study causal relations of trait and state variables. The hypothesis is that traits should cause states, which in turn cause behavior.

Subtopic #3

Based on results of Subtopic #1 and #2, develop algorithms that can be generalized to different datasets for the purpose of extracting trait/state like variables.

Project #2 (complete)

I worked on two-photon calcium imaging data of the primary visual cortex (V1) of an awake macaque monkey. I used VGG19 neural network and Imagenet images to study the logics behind that only a few neurons respond highly to encode each image the monkey was seeing (poster).

Project #3 (complete)

The goal was to study individual neurons’ stimulus selectivity and their functional connectivity. The data I used was a large-scale dataset recorded by two-photon calcium imaging in the primary visual cortex (V1) of an awake macaque monkey. I studied what are the favorite stimuli of each neuron in the recorded group of about 1000 neurons, using Gaussian models. I analyzed the identities of neurons, such as curve-tuning neurons are encoding curvature of different curves seen by the monkey, and angle-tuning neurons are encoding opening degree of different angles (publication).

Viewing each neuron as a node and each edge representing the functional connectivity between the neurons, I then used graphical models to study network dynamics of the neurons. I studied how the neurons are clustered differently when encoding different images.

Home / Project