Extreme Learning to Rank via Low Rank Assumption

Authors: Minhao Cheng, Ian Davidson, Cho-Jui Hsieh

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

Reproducibility Variable Result LLM Response
Research Type Experimental We conduct experiments on real world datasets, showing that the algorithm achieves higher accuracy and faster training time compared with state-of-the-art methods.
Researcher Affiliation Academia 1Department of Computer Science, University of California Davis, USA. 2Department of Statistics, University of California Davis, USA.
Pseudocode Yes Algorithm 1 Factorization Rank SVM: Computing Uf(U, V )
Open Source Code No The paper does not provide concrete access to source code for the methodology described.
Open Datasets Yes We choose three datasets in our real-world application experiments: (1) Yahoo! Movies User Ratings and Descriptive Content Information V1 01 (2) Het Rec2011-Movie Lens-2K (Cantador et al., 2011). (3) Movie Lens 20M Dataset (Harper & Konstan, 2016).
Dataset Splits Yes For each ranking task, we randomly split the items into training items and testing items. In the training phase, we use all the pairs between training items to train the model, and in the testing phase we evaluate the prediction accuracy for all the testing-testing item pairs and testing-training item pairs, which is similar with BPR (Rendle et al., 2009). ... We choose the best regularization parameter for each method by a validation set.
Hardware Specification Yes All the experiments are conducted on a server with an Intel E7-4820 CPU and 256G memory.
Software Dependencies No The paper mentions software components like 'square hinge loss' and 'gradient descent' but does not provide specific software names with version numbers for reproducibility.
Experiment Setup Yes We choose the best regularization parameter for each method by a validation set.