ATP: Directed Graph Embedding with Asymmetric Transitivity Preservation

Authors: Jiankai Sun, Bortik Bandyopadhyay, Armin Bashizade, Jiongqian Liang, P. Sadayappan, Srinivasan Parthasarathy265-272

AAAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments show that our framework consistently and significantly outperforms the state-of-the-art baselines on three diverse realworld tasks: link prediction, and question difficulty estimation and expert finding in online forums like Stack Exchange.
Researcher Affiliation Academia Jiankai Sun, Bortik Bandyopadhyay, Armin Bashizade, Jiongqian Liang, P. Sadayappan, Srinivasan Parthasarathy Department of Computer Science and Engineering, The Ohio State University, USA {sun.1306,bandyopadhyay.14,bashizade.1,liang.420,sadayappan.1}.osu.edu,srini@cse.ohio-state.edu
Pseudocode No The paper describes the proposed methodology verbally in sections such as 'Our Framework ATP' but does not include any structured pseudocode or algorithm blocks.
Open Source Code No The paper does not contain any explicit statements about open-source code availability for the described methodology, nor does it provide any links to a code repository.
Open Datasets Yes Datasets used for evaluation are Wiki-Vote, GNU, Cit Hep PH, which were used in prior work (Lai et al. 2017). https://snap.stanford.edu/data/wiki-Vote.html ... We used the data dump which is released on June 12, 2017 and is available online at https://archive.org/details/stackexchange
Dataset Splits No The paper describes how positive examples are selected for link prediction ('We select r = 10% edges as positive examples.') and refers to prior work for cold question selection ('We followed the same settings proposed by Jiankai et al. (Sun et al. 2018b) to select cold questions for evaluation'). However, it does not explicitly provide specific training/validation/test dataset splits (e.g., percentages or absolute counts) for the overall experiments.
Hardware Specification No The paper does not provide any specific details regarding the hardware (e.g., CPU, GPU models, memory, or cloud instances) used to conduct the experiments.
Software Dependencies No The paper mentions using 'a multi-core GPU version of NMF using CCD GPUCCD++' but does not provide specific version numbers for this or any other software dependencies such as programming languages or libraries.
Experiment Setup Yes To predict whether there is a directed path from node i to node j, we check the value of σ( si, tj ), where σ is the sigmoid function, si is node i s source vector and tj is node j s target vector respectively. If σ( si, tj ) > α, there is a predicted path from i to j. α is a threshold with range in [0.5, 1). We set α = 0.5 in our experiments, which we empirically found to work well.