Similarity-Navigated Conformal Prediction for Graph Neural Networks

Authors: Jianqing Song, Jianguo Huang, Wenyu Jiang, Baoming Zhang, Shuangjie Li, Chongjun Wang

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

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments demonstrate the superiority of SNAPS, improving the efficiency of prediction sets and singleton hit ratio while maintaining valid coverage.
Researcher Affiliation Academia 1State Key Laboratory of Novel Software Technology, Nanjing University 2Department of Statistics and Data Science, Southern University of Science and Technology 3College of Computing and Data Science, Nanyang Technological University
Pseudocode Yes The pseudo-code for SNAPS is presented in Algorithm 1.
Open Source Code Yes Code is available at https://github.com/janqsong/SNAPS.
Open Datasets Yes We conduct thorough empirical evaluations on 10 datasets, including both small datasets and large-scale datasets, e.g., OGBN Products (Bhatia et al., 2016).
Dataset Splits Yes Vlabel is then randomly split into Vtrain/Vvalid/Vcalib with a fixed size, the training/validation/calibration node set, correspondingly.
Hardware Specification Yes Each experiment is done with a single NVIDIA V100 32GB GPU.
Software Dependencies No The paper mentions using 'PyTorch Geometric' and 'torchvision repository' but does not specify software versions for these or other dependencies like Python or PyTorch itself.
Experiment Setup Yes For SNAPS, we choose "lambda" and "mu" in increments of 0.05 within the range 0 to 1, and ensure that "lambda" + "mu" <= 1.