zeigar / tmkl

Resting State Dynamics Meets Anatomical Structure: Temporal Multiple Kernel Learning (tMKL) Model

tMKL

Resting State Dynamics Meets Anatomical Structure: Temporal Multiple Kernel Learning (tMKL) Model

Dependencies

python3 sklearn

Get Started

Open matlab and run the following:

run <tMKL_dir_path>/CODE/main_script.m

Directory Description

1. CODE - contains scripts and functions

• main_script.m - main script for running tMKL
• get_train_struct.m - wFCs generation, vectorize wFCs, spectral embedding - projection to manifold, GMM - assign wFCs to states
• run_tMKL.m - run MKL model for every state
• grand_average_prediction.m - evaluate model performance

2. INPUT_MATS - stores the data, including SC FC matrices and BOLD timeseries data.
3. INTER_MATS - stores intermediate representations : wFCs, cluster assignments, transition matrices, tMKL model parameters etc.
4. OUTPUT_MATS - stores final model performance results