The code implements a family of Concept Graph Learning (CGL) algorithms developed in the following papers:

Hanxiao Liu, Wanli Ma, Yiming Yang, and Jaime Carbonell. "Learning Concept Graphs from Online Educational Data." In Journal of Artificial Intelligence Research 55 (2016): 1059-1090. [PDF]

Yiming Yang, Hanxiao Liu, Jaime Carbonell, and Wanli Ma. "Concept graph learning from educational data." In the Eighth ACM International Conference on Web Search and Data Mining, pp. 159-168. ACM, 2015. [PDF]

More details about the task and datasets can be found at our project webpage. Please cite the above papers if you end up using our code/datasets.

Concept graph automatically induced from MIT OpenCourseWare:

To conduct cross-validation using plain CGL, run

matlab -r main

Configurations of the program are located at config.m. To allow graph transduction, set

opt.transductive = false;

To carry out sparse CGL, set

opt.algorithm = @cgl_rank_sparse;

Hanxiao Liu, Carnegie Mellon University.