Arbitrary Conditional Distributions with Energy

Authors: Ryan Strauss, Junier B. Oliva

NeurIPS 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We evaluate ACE on benchmark datasets and show that it outperforms current methods for arbitrary conditional/marginal density estimation. ACE remains effective when trained on data with missing values, making it applicable to real-world datasets that are often incomplete, and we find that ACE scales well to high-dimensional data. Also, unlike some prior methods (e.g., [20]), ACE can naturally model data with both continuous and discrete values.
Researcher Affiliation Academia Ryan R. Strauss Department of Computer Science UNC at Chapel Hill Chapel Hill, NC 27514 rrs@cs.unc.edu Junier B. Oliva Department of Computer Science UNC at Chapel Hill Chapel Hill, NC 27514 joliva@cs.unc.edu
Pseudocode Yes The pseudocode for this procedure is presented in the Appendix (see Algorithm 1).
Open Source Code Yes An implementation of ACE is available at https://github.com/lupalab/ace.
Open Datasets Yes We first evaluate ACE on real-valued tabular data. Specifically, we consider the benchmark UCI repository datasets described by Papamakarios et al. [25] (see Table 6 in the Appendix). ... The processed data has 6 continuous features and 8 discrete features and is split into train, validation, and test partitions of size 22003, 5501, and 13788 respectively.
Dataset Splits Yes The processed data has 6 continuous features and 8 discrete features and is split into train, validation, and test partitions of size 22003, 5501, and 13788 respectively.
Hardware Specification No The paper does not provide specific details about the hardware used for running experiments, such as GPU models, CPU types, or cloud instance specifications.
Software Dependencies No The paper does not provide specific version numbers for ancillary software dependencies or libraries, only mentioning the optimizer Adam.
Experiment Setup No Full experimental details and hyperparameters can be found in the Appendix.