Relaxing Bijectivity Constraints with Continuously Indexed Normalising Flows

Authors: Rob Cornish, Anthony Caterini, George Deligiannidis, Arnaud Doucet

ICML 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We evaluated the performance of CIFs on several problems of varying difficulty, including synthetic 2-D data, several tabular datasets, and three image datasets.
Researcher Affiliation Academia 1University of Oxford, Oxford, United Kingdom 2The Alan Turing Institute, London, United Kingdom. Correspondence to: Rob Cornish <rcornish@robots.ox.ac.uk>.
Pseudocode Yes Algorithm 1 Unbiased estimation of L(x)
Open Source Code Yes See github. com/jrmcornish/cif for our code.
Open Datasets Yes We tested the performance of CIFs on the tabular datasets used by Papamakarios et al. (2017). For each dataset, we trained 10 and 100-layer baseline fully connected Res Flows, and corresponding 10-layer CIF-Res Flows... We also considered CIFs applied to the MNIST (Le Cun, 1998), Fashion-MNIST (Xiao et al., 2017), and CIFAR-10 (Krizhevsky & Hinton, 2009) datasets.
Dataset Splits No The paper mentions training and testing but does not explicitly provide details about a validation dataset split or its use.
Hardware Specification No The paper does not provide specific hardware details (e.g., GPU/CPU models, memory) used for running its experiments.
Software Dependencies No The paper does not provide specific software dependencies with version numbers (e.g., library names, framework versions) needed to replicate the experiment.
Experiment Setup No Full experimental details, including additional 2-D figures along the lines of Figure 1, are in Section C of the Supplement.