Group-sparse Embeddings in Collective Matrix Factorization

Authors: Arto Klami; Guillaume Bouchard; Abhishek Tripathi

ICLR 2014 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We compare MAP and variational Bayesian solutions based on alternating optimization algorithms and show that the model automatically infers the nature of each factor using group-wise sparsity. Our approach supports in a principled way continuous, binary and count observations and is efficient for sparse matrices involving missing data. We illustrate the solution on a number of examples, focusing in particular on an interesting use-case of augmented multi-view learning.
Researcher Affiliation Collaboration Arto Klami arto.klami@cs.helsinki.fi Helsinki Institute for Information Technology HIIT, Department of Information and Computer Science, University of Helsinki Guillaume Bouchard guillaume.bouchard@xrce.xerox.com Xerox Research Centre Europe Abhishek Tripathi abishek.tripathi3@xerox.com Xerox Research Centre India
Pseudocode Yes The full algorithm repeats the following steps until convergence.
Open Source Code No The paper does not provide concrete access to source code (e.g., specific repository link, explicit code release statement, or code in supplementary materials) for the methodology described.
Open Datasets Yes We start with a multi-view setup in computational biology, using data from Pollack et al. (2002) and the setup studied by Klami et al. (2013). ... Next we consider classical recommender systems, using Movie Lens and Flickr data as used in earlier CMF experiments by Bouchard et al. (2013).
Dataset Splits Yes For MAP we validate the strength of the Gamma hyper-priors for τ and α over a grid of 11 × 11 values for a0 = b0 and p0 = q0, using two-fold cross-validation within the observed data.
Hardware Specification No The paper states: "Both data sets have roughly 1 million observed entries, and our solutions were computed in a few minutes on a laptop." This provides a general type of device but lacks specific hardware details such as CPU model, GPU, or memory capacity.
Software Dependencies No The paper does not provide specific ancillary software details (e.g., library or solver names with version numbers) needed to replicate the experiment.
Experiment Setup Yes For all methods we use sufficiently large K, letting ARD prune out unnecessary components, and run the algorithms until the variational lower bound converges. ... For MAP we validate the strength of the Gamma hyper-priors for τ and α over a grid of 11 × 11 values for a0 = b0 and p0 = q0, using two-fold cross-validation within the observed data.