Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in [1].

Bayesian Inference with Posterior Regularization and Applications to Infinite Latent SVMs

Authors: Jun Zhu, Ning Chen, Eric P. Xing

JMLR 2014 | Venue PDF | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We present efficient inference methods and report empirical studies on several benchmark data sets, which appear to demonstrate the merits inherited from both large-margin learning and Bayesian nonparametrics.
Researcher Affiliation Academia Jun Zhu EMAIL Department of Computer Science and Technology State Key Laboratory of Intelligent Technology and Systems Tsinghua National Laboratory for Information Science and Technology Tsinghua University Beijing, 100084 China Ning Chen EMAIL Department of Computer Science and Technology State Key Laboratory of Intelligent Technology and Systems Tsinghua National Laboratory for Information Science and Technology Tsinghua University Beijing, 100084 China Eric P. Xing EMAIL School of Computer Science Carnegie Mellon University Pittsburgh, PA 15213, USA
Pseudocode Yes Algorithm 1 Inference Algorithm for Infinite Latent SVMs
Open Source Code No The paper does not provide concrete access to source code for the methodology described. It only provides a link to data splits for a specific dataset, not its own implementation.
Open Datasets Yes We evaluate the infinite latent SVM (i LSVM) for classification on the real TRECVID2003 and Flickr image data sets... The Flickr image data set... downloaded from the Flickr website.13 The website is available at: http://www.flickr.com/. These data sets are from the UCI repository, and each data example has multiple labels. The School data set... It consists of examination records... The data set is publicly available and has been extensively evaluated in various multi-task learning methods...
Dataset Splits Yes We follow the same training/testing splits as in Chen et al. (2012). The Yeast data set consists of 1500 training and 917 test examples... The Scene data set consists 1211 training and 1196 test examples... We use the same 10 random splits16 of the data, so that 75% of the examples from each school (task) belong to the training set and 25% to the test set.
Hardware Specification No The paper does not provide specific hardware details (e.g., exact GPU/CPU models, processor types, or memory amounts) used for running its experiments.
Software Dependencies No These dual problems (or their primal forms) can be efficiently solved with a binary SVM solver, such as SVM-light or Lib SVM. No specific version numbers for these software tools are provided.
Experiment Setup Yes We perform 5-fold cross-validation on training data to select hyperparameters, e.g., α and C (we use the same procedure for MT-i LSVM). Initialize γk1 = α, γk2 = 1, ψdk = 0.5 + ϵ, where ϵ N(0, 0.001), Φmn = 0, σ2 mn = σ2 m0 = 1, µm = 0, λ2 mn is computed from D. until relative change of L is less than τ (e.g., 1e-3) or iteration number is T (e.g., 10)