BooVAE: Boosting Approach for Continual Learning of VAE

Authors: Evgenii Egorov, Anna Kuzina, Evgeny Burnaev

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

Reproducibility Variable Result LLM Response
Research Type Experimental We empirically validate the proposed algorithm on commonly used benchmarks (MNIST, Fashion-MNIST, Not MNIST) and Celeb A for disjoint sequential image generation tasks.
Researcher Affiliation Academia Evgenii Egorov University of Amsterdam egorov.evgenyy@ya.ru Anna Kuzina Vrije Universiteit av.kuzina@yandex.ru Evgeny Burnaev Skoltech, AIRI e.burnaev@skoltech.ru
Pseudocode Yes Algorithm 1 Boo VAE algorithm
Open Source Code Yes We provide code at https://github.com/AKuzina/BooVAE.
Open Datasets Yes We perform experiments on MNIST, not MNIST, fashion MNIST and Celeb A datasets.
Dataset Splits No The paper mentions using a 'test dataset' and a 'training dataset' for experiments but does not explicitly provide specific percentages or counts for training, validation, and test splits.
Hardware Specification Yes In Supp.(B.5) we mention that we use 4 NVIDIA V100 GPU for each experiment.
Software Dependencies No The paper mentions PyTorch in relation to Inception V3 network but does not explicitly list other software dependencies with specific version numbers in the main text.
Experiment Setup Yes For each task, we add a new classification head (one fully connected layer) and train for 200 epochs with batch size 500.