Collaborative Cognitive Diagnosis with Disentangled Representation Learning for Learner Modeling

Authors: Weibo Gao, Qi Liu, Linan Yue, Fangzhou Yao, Hao Wang, Yin Gu, zheng zhang

NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments demonstrate the superior performance of Coral, showcasing significant improvements over state-of-the-art methods across several real-world datasets.
Researcher Affiliation Academia State Key Laboratory of Cognitive Intelligence, University of Science and Technology of China
Pseudocode Yes Algorithm 1 Coral Model
Open Source Code Yes Our code is available at https://github.com/bigdata-ustc/Coral.
Open Datasets Yes We conduct experiments on three real-world datasets: ASSIST [11], Junyi [5] and Neur IPS2020EC [43].
Dataset Splits Yes In this setting, we split all the datasets with a 7:1:2 ratio into training sets, validation sets, and test sets.
Hardware Specification Yes All experiments are conducted on a Linux server equipped with two 3.00GHz Intel Xeon Gold 5317 CPUs and two Tesla A100 40G GPUs.
Software Dependencies No Each model is implemented by Py Torch [37] and optimized by Adam optimizer [19]. (No version numbers specified for PyTorch or Adam)
Experiment Setup Yes We set the dimension size d as 20, the layer of graph modeling as 2, and the mini-batch size as 512. In the training stage, we select the learning rate from {0.002, 0.005, 0.01, 0.02, 0.05}, select α from {0.05, 0.1, 0.5, 1} and β from {0.25, 0.5, 1}, and select neighboring number K from {1, 2, 3, 4, 5, 10, 15, 20, 15, 30, 25, 40, 45, 50}. All network parameters are initialized with Xavier initialization [15].