Continual Unsupervised Disentangling of Self-Organizing Representations

Authors: Zhiyuan Li, Xiajun Jiang, Ryan Missel, Prashnna Kumar Gyawali, Nilesh Kumar, Linwei Wang

ICLR 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We tested the presented method on a split version of 3DShapes to provide the quantitative disentanglement evaluation of continually learned representations, and further demonstrated its ability to continually disentangle new representations and improve shared downstream tasks in benchmark datasets.
Researcher Affiliation Academia Zhiyuan Li1, Xiajun Jiang1, Ryan Missel1, Prashnna Kumar Gyawali2, Nilesh Kumar1, Linwei Wang1 Rochester Institute of Technology1, Stanford University2
Pseudocode No The paper does not contain explicit pseudocode or algorithm blocks.
Open Source Code Yes Code is available at https://github.com/Zhiyuan1991/CUDOS_release.
Open Datasets Yes We evaluated CUDOS on (1) a split version of 3DShapes (Burgess & Kim, 2018)... (2) MNIST(Le Cun et al., 1998), Fashion-MNIST(Xiao et al., 2017), and their moving versions in (Achille et al., 2018), and (3) split-Celeb A (Liu et al., 2015).
Dataset Splits No The paper states that it uses 'a split version of 3DShapes' and other datasets, but it does not provide specific train/validation/test split percentages, sample counts, or references to predefined splits for reproduction.
Hardware Specification No The paper does not explicitly mention the specific hardware (e.g., GPU models, CPU types, memory) used for running the experiments.
Software Dependencies No The paper provides hyperparameters but does not list specific software dependencies with version numbers (e.g., 'Python 3.8', 'PyTorch 1.9').
Experiment Setup Yes We set γ1 = 0.25, γ2 = 1, γ3 = 0.35, b = 10 in all experiments. Snapshot of the model is updated every τ = 1500 iteration steps.