Dyadic Prediction Using a Latent Feature Log-Linear Model - Computer Science > LearningReport as inadecuate




Dyadic Prediction Using a Latent Feature Log-Linear Model - Computer Science > Learning - Download this document for free, or read online. Document in PDF available to download.

Abstract: In dyadic prediction, labels must be predicted for pairs dyads whosemembers possess unique identifiers and, sometimes, additional features calledside-information. Special cases of this problem include collaborative filteringand link prediction. We present the first model for dyadic prediction thatsatisfies several important desiderata: i labels may be ordinal or nominal,ii side-information can be easily exploited if present, iii with or withoutside-information, latent features are inferred for dyad members, iv it isresistant to sample-selection bias, v it can learn well-calibratedprobabilities, and vi it can scale to very large datasets. To our knowledge,no existing method satisfies all the above criteria. In particular, manymethods assume that the labels are ordinal and ignore side-information when itis present. Experimental results show that the new method is competitive withstate-of-the-art methods for the special cases of collaborative filtering andlink prediction, and that it makes accurate predictions on nominal data.



Author: Aditya Krishna Menon, Charles Elkan

Source: https://arxiv.org/







Related documents