PROGRESSIVE LEARNING AND DISENTANGLEMENT OF HIERARCHICAL REPRESENTATIONS

Authors: Zhiyuan Li, Jaideep Vitthal Murkute, Prashnna Kumar Gyawali, Linwei Wang

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

Reproducibility Variable Result LLM Response
Research Type Experimental We quantitatively demonstrate the ability of the presented model to improve disentanglement in comparison to existing works on two benchmark data sets using three disentanglement metrics, including a new metric we proposed to complement the previously-presented metric of mutual information gap.
Researcher Affiliation Academia Zhiyuan Li, Jaideep Vitthal Murkute, Prashnna Kumar Gyawali & Linwei Wang Golisano College of Computing and Information Sciences Rochester Institute of Technology Rochester, NY 14623, USA {zl7904,jvm6526,pkg2182,Linwei.Wang}@rit.edu
Pseudocode No The paper describes the model and methods in prose and mathematical equations but does not include any distinct pseudocode or algorithm blocks.
Open Source Code Yes Source code available at https://github.com/Zhiyuan1991/proVLAE.
Open Datasets Yes We tested the presented pro-VLAE on four benchmark data sets: d Sprites (Matthey et al. (2017)), 3DShapes (Burgess & Kim (2018)), MNIST (Le Cun et al. (1998)), and Celeb A (Liu et al. (2015)), where the first two include ground-truth generative factors that allow us to carry out comprehensive quantitative comparisons of disentangling metrics with existing models.
Dataset Splits No The paper mentions training epochs and datasets, but does not provide specific train/validation/test split percentages, sample counts, or explicit splitting methodology needed to reproduce the data partitioning. It uses terms like 'train' in the context of the model, but not for dataset splits explicitly.
Hardware Specification No The paper does not provide specific hardware details (e.g., GPU/CPU models, processor types, memory amounts, or detailed computer specifications) used for running its experiments.
Software Dependencies No The paper does not provide specific ancillary software details (e.g., library or solver names with version numbers) needed to replicate the experiment.
Experiment Setup Yes In specific, we introduce a fade-in coefficient α to equations (6) and (7) when growing new components in the encoder and the decoder: [...] where α increases from 0 to 1 within a certain number of iterations (5000 in our experiments) since the addition of the new network components µl,σl, and ml. [...] where γ is set to 0.5 in our experiments.