Towards Open-World Feature Extrapolation: An Inductive Graph Learning Approach

Authors: Qitian Wu, Chenxiao Yang, Junchi Yan

NeurIPS 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our experiments over several classification datasets and large-scale advertisement click prediction datasets demonstrate that our model can produce effective embeddings for unseen features and significantly outperforms baseline methods that adopt KNN and local aggregation.
Researcher Affiliation Academia Qitian Wu, Chenxiao Yang, Junchi Yan Department of Computer Science and Engineering Mo E Key Lab of Artificial Intelligence, AI Institute Shanghai Jiao Tong University {echo740, chr26195, yanjunchi}@sjtu.edu.cn
Pseudocode Yes We present the complete training algorithm in Appendix B where the model is trained end-to-end using self-supervised or inductive learning approaches.
Open Source Code Yes The implementation codes are public available at https://github.com/qitianwu/FATE.
Open Datasets Yes First, we consider six classification datasets from UCI Machine Learning Repository [1]: Gene, Protein, Robot, Drive, Calls and Github, as collected from domains like biology, engineering and social domains with diverse features, as well as two large-scale datasets Avazu and Criteo from real-world online advertisement system whose goal is to predict the Click-Through Rate (CTR) of exposed advertisement to users.
Dataset Splits Yes We randomly split all the instances into training/validation/test data with the ratio 6:2:2.
Hardware Specification No The paper does not provide specific hardware details (such as exact GPU/CPU models, processor types, or memory amounts) used for running its experiments.
Software Dependencies No The paper mentions that the model is implemented in Python and PyTorch, but does not provide specific version numbers for these or any other ancillary software components.
Experiment Setup Yes We specify FATE in the following ways. 1) Backbone: a 3-layer feedforward NN. 2) GNN: a 4-layer GCN. 3) Training: self-supervised learning with n-fold splits.