Bayesian Matrix Completion via Adaptive Relaxed Spectral Regularization

Authors: Yang Song, Jun Zhu

AAAI 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments on synthetic and real datasets demonstrate encouraging results on rank recovery and collaborative filtering, with notably good results for very sparse matrices.
Researcher Affiliation Academia Yang Song and Jun Zhu Department of Physics, Tsinghua University, yang.song@zoho.com Department of Computer Science & Tech., State Key Lab of Intell. Tech. & Sys.; CBICR Center; Tsinghua National Lab for Information Science and Tech., Tsinghua University, dcszj@mail.tsinghua.edu.cn
Pseudocode No No structured pseudocode or algorithm blocks were found. The paper describes the inference steps in prose.
Open Source Code No The paper does not provide an explicit statement about releasing source code for the described methodology or a link to a code repository. Footnote 1 links to supplementary material, not code.
Open Datasets Yes Movie Lens 1M2 and Each Movie datasets, and compare results with various strong competitors... 2Movie Lens datasets can be downloaded from http://grouplens.org/datasets/movielens/.
Dataset Splits Yes We randomly split the dataset into 80% training and 20% test. We further split 20% training data for validation for M3F, i PM3F, Soft Impute, Soft Impute-ALS and HASI to tune their hyperparameters.
Hardware Specification No No specific hardware details (e.g., CPU/GPU models, memory) used for running the experiments are provided in the paper.
Software Dependencies No The paper mentions 'R package soft Impute' for Soft Impute and Soft Impute-ALS, and states using 'the code provided by the corresponding authors' for other methods, but does not provide specific version numbers for any software dependencies.
Experiment Setup Yes For each matrix Z, the iteration number was fixed to 1000 and the result was averaged from last 200 samples (with first 800 discarded as burn-in). We simply initialize our sampler with uniformly distributed U and V with norms fixed to 0.9 and all d fixed to zero.