Subgraph Pattern Neural Networks for High-Order Graph Evolution Prediction

Authors: Changping Meng, S Chandra Mouli, Bruno Ribeiro, Jennifer Neville

AAAI 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Empirical results show that our proposed method significantly outperforms other stateof-the-art methods designed for static and/or single node/link prediction tasks.
Researcher Affiliation Academia Changping Meng, S Chandra Mouli, Bruno Ribeiro, Jennifer Neville Department of Computer Science Purdue University, West Lafayette, IN {meng40, chandr}@purdue.edu, {ribeiro, neville}@cs.purdue.edu
Pseudocode No The paper describes the model architecture and steps in detail but does not include any explicitly labeled 'Pseudocode' or 'Algorithm' blocks.
Open Source Code Yes Source code is available at https://github.com/PurdueMINDS/SPNN.
Open Datasets Yes We use two representative heterogeneous graph datasets with temporal information. DBLP (Sun et al. 2011b) contains scientific papers... Facebook is a sample of the Facebook users from one university. ... Word Net is a knowledge graph that groups words into synonyms... The WN18 dataset is a subset of Word Net...
Dataset Splits Yes 20% of the training examples are separated as validation for early stopping.
Hardware Specification Yes The server is an Intel E5 2.60GHz CPU with 512 GB of memory.
Software Dependencies No The paper states 'We implement SPNN in Theano' but does not provide version numbers for Theano or any other software dependencies.
Experiment Setup Yes The loss function is the negative log likelihood plus L1 and L2 regularization penalties over the parameters, both with regularization penalty 0.001. We train SPNN using stochastic gradient descent over a maximum of 30000 epochs and learning rate 0.01.