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].

Masked Autoregressive Flow for Density Estimation

Authors: George Papamakarios, Theo Pavlakou, Iain Murray

NeurIPS 2017 | Venue PDF | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We experimentally evaluate MAF on a wide range of datasets, and we demonstrate that (a) MAF outperforms Real NVP on general-purpose density estimation, and (b) a conditional version of MAF achieves close to state-of-the-art performance on conditional image modelling even with a general-purpose architecture.
Researcher Affiliation Academia George Papamakarios University of Edinburgh EMAIL Theo Pavlakou University of Edinburgh EMAIL Iain Murray University of Edinburgh EMAIL
Pseudocode No The paper describes algorithms and transformations using mathematical equations and descriptive text, but does not include explicit pseudocode or algorithm blocks.
Open Source Code Yes Code for reproducing our experiments (which uses Theano [29]) can be found at https://github.com/gpapamak/maf.
Open Datasets Yes Following Uria et al. [32], we perform unconditional density estimation on four UCI datasets (POWER, GAS, HEPMASS, MINIBOONE) and on a dataset of natural image patches (BSDS300). ... For conditional density estimation, we used the MNIST dataset of handwritten digits [17] and the CIFAR-10 dataset of natural images [14].
Dataset Splits Yes Each model was trained with early stopping until no improvement occurred for 30 consecutive epochs on the validation set. For each model, we selected the number of hidden layers and number of hidden units based on validation performance
Hardware Specification No The paper mentions "parallel computing architectures such as Graphics Processing Units (GPUs)" but does not specify any particular GPU models or other hardware details used for the experiments.
Software Dependencies No The paper mentions using "Theano [29]" but does not specify its version number or any other software dependencies with their versions.
Experiment Setup Yes All models were trained with the Adam optimizer [11], using a minibatch size of 100, and a step size of 10 3 for MADE and MADE Mo G, and of 10 4 for Real NVP and MAF. A small amount of โ„“2 regularization was added, with coef๏ฌcient 10 6.