Learning Binary Latent Variable Models: A Tensor Eigenpair Approach

Authors: Ariel Jaffe, Roi Weiss, Boaz Nadler, Shai Carmi, Yuval Kluger

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

Reproducibility Variable Result LLM Response
Research Type Experimental We demonstrate our method in three scenarios: (I) simulations from the exact binary model (1); (II) learning a common population genetic admixture model; (III) learning the proportion matrix of a cell mixture from DNA methylation levels.
Researcher Affiliation Academia 1Dept. of Computer Science and Applied Mathematics, Weizmann Institute of Science, Rehovot 7610001, Israel. 2Braun School of Public Health and Community Medicine, The Hebrew University of Jerusalem, Jerusalem 9112102, Israel. 3Program of Applied Mathematics, Yale University, New Haven, CT 06511, USA.
Pseudocode Yes Algorithm 1 Estimate W when σ > 0 and n <
Open Source Code Yes Code to reproduce the simulation results can be found at https://github.com/ar Jaffe/ Binary Latent Variables.
Open Datasets No The paper describes generating 'n samples from model (1)' for simulations and 'simulating genetic admixture' using specific schemes and parameters. It does not refer to or provide access information for any pre-existing, publicly available datasets for training purposes.
Dataset Splits No The paper discusses generating samples for experiments and using a sample split (one part for computing candidates, another for filtering), but it does not explicitly specify traditional training/validation/test splits with percentages or sample counts for model development or evaluation.
Hardware Specification No The paper states: 'Practically, our method, implemented without any particular optimization, can learn a model with 12 hidden units in less than ten minutes on a standard PC.' The term 'standard PC' is too general and does not provide specific hardware details like CPU/GPU models or memory.
Software Dependencies No The paper mentions 'With our current Matlab implementation', but it does not specify any version numbers for Matlab or any other software libraries or dependencies used in the experiments.
Experiment Setup Yes We generated n samples from model (1) with d = 6 hidden units, m = 30 observable features, and Gaussian noise ξ N(0, Im). The m columns of W were drawn uniformly from the unit sphere Sd 1. Fixing a mean vector a Rd and a covariance matrix R Rd d, each hidden vector h was generated independently by first drawing r N(a, R) and then taking its binary rounding.