Covariance Modeling with external variates
Data mining using Canonical Correlation Analysis for Time Series
- In this project, we introduce the time-dependent canonical correlation analysis (TDCCA), a method of inferring time-dependent canonical vectors of paired variables.
Spatial and temporal correlation analysis with application of fMRI data
- fMRI data is usually temporal correlated. In this project, the temporal correlation is considered to improve spatial correlation in the setting of Canonical Correlation Analysis.
Estimating high dimensional ODE models from convoluted observations with an application to fMRI
- It aims to provide a sparse dynamic network estimation not only for fMRI data but for other possible data that can well be represented by convolution model.
Causal Dynamic Network Analysis
- It aims to improve the dynamic causal modeling with the optimization based method. Our optimization-based method and algorithm compute efficiently the ODE parameters from fMRI data, instead of comparing potentially a huge number of candidate ODE models.
Spatial Reasoning with Deep Learning
- This project focuses on the study of whether cutting-edge deep learning methods can beat the human on the difficult task like spatial reasoning which is very important to many AI techniques including autonomous cars.