A novel deep learning model for collective classification in multi-relational domains.
A machine learning model normally learns from data with an assumption that all the instances are iid. However, ignoring the dependency among instances may hurt the performance in many data. For example, in a citation network, two publications may be in the same category if one cites the other. Our work models multiple-type relations among instances in data using neural networks. This improves the classification performance in 3 networked datasets.
Link to the paper: https://arxiv.org/abs/1609.04508
See training_instruction.txt for running experiments