Large-Scale Heterogeneous Feature Embedding

Authors: Xiao Huang, Qingquan Song, Fan Yang, Xia Hu3878-3885

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

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments on four real-world datasets demonstrate the efficiency and effectiveness of Feat Walk.
Researcher Affiliation Academia Department of Computer Science and Engineering, Texas A&M University {xhuang, song 3134, nacoyang, xiahu}@tamu.edu
Pseudocode No The paper describes the steps of the 'Feature Walks' process in prose, but it does not include any clearly labeled pseudocode or algorithm blocks.
Open Source Code No The paper does not provide an explicit statement about the availability of the source code for the described methodology, nor does it provide a link to a code repository.
Open Datasets Yes The four real-world datasets that we employed in the experiments are all publicly available. Reuters (Amini, Usunier, and Goutte 2009), Flickr (Huang, Li, and Hu 2017b), ACM (Tang et al. 2008), Yelp (Yelp 2017): www.yelp.com/dataset/challenge.
Dataset Splits Yes We apply 5-fold cross-validation on all datasets, i.e., randomly select 4/5 of all instances as a training group and the remaining as a test group. We vary it as {25%, 50%, 100%}.
Hardware Specification Yes We ran the experiments on a Dell Opti Plex 9030 i7-16GB desktop.
Software Dependencies No The paper mentions using word2vec and an SVM classifier (referencing Pedregosa et al. 2011 for scikit-learn) but does not provide specific version numbers for these or any other software dependencies.
Experiment Setup Yes If it is not specified, d is set as 100 and 100% of the instances in the training group are used. We vary α from 0 to 1 and the window size from 1 to 10. We vary the walk length L as from 2 to 60 and the number of sentences per instance W/N from 2 to 20. We vary d as {20, 60, 100, 140, 180}.