Discretely Relaxing Continuous Variables for tractable Variational Inference

Authors: Trefor Evans, Prasanth Nair

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

Reproducibility Variable Result LLM Response
Research Type Experimental We empirically demonstrate our approaches in section 5...Our empirical studies demonstrate superior performance relative to competing SVI methods on problems with as many as 2 million training points. In this section, we consider discretely relaxing a continuous Gaussian prior on the weights of a generalized linear regression model. This allows us to compare performance between a reparameterization gradient estimator for a continuous prior and our DIRECT method for a relaxed, discrete prior. Considering regression datasets from the UCI repository, we report the mean and standard deviation of the root mean squared error (RMSE) from 10-fold cross validation... In table 1, we see the results of our studies across several model-types.
Researcher Affiliation Academia Trefor W. Evans University of Toronto trefor.evans@mail.utoronto.ca Prasanth B. Nair University of Toronto pbn@utias.utoronto.ca
Pseudocode No The paper does not contain structured pseudocode or algorithm blocks.
Open Source Code Yes Our full code is available at https://github.com/treforevans/direct.
Open Datasets Yes Considering regression datasets from the UCI repository, we report the mean and standard deviation of the root mean squared error (RMSE) from 10-fold cross validation. Also presented is the mean training time per fold on a machine with two E5-2680 v3 processors and 128Gb of RAM, and the expected sparsity (percentage of zeros) within a posterior sample. All models use b 2000 basis functions. Further details of the experimental setup can be found in appendix E. In table 1, we see the results of our studies across several model-types.
Dataset Splits Yes we report the mean and standard deviation of the root mean squared error (RMSE) from 10-fold cross validation... 390% train, 10% test per fold.
Hardware Specification Yes Also presented is the mean training time per fold on a machine with two E5-2680 v3 processors and 128Gb of RAM
Software Dependencies No The paper mentions 'Tensor Flow' but does not specify a version number or other software dependencies with version information.
Experiment Setup Yes Further details of the experimental setup can be found in appendix E.