Topic-VQ-VAE: Leveraging Latent Codebooks for Flexible Topic-Guided Document Generation

Authors: YoungJoon Yoo, JongWon Choi

AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our experimental results on document analysis and image generation demonstrate that TVQ-VAE effectively captures the topic context which reveals the underlying structures of the dataset and supports flexible forms of document generation." and "Empirical Analysis We analyze the TVQ-VAE performance with two different applications: document analysis and image generation.
Researcher Affiliation Collaboration Young Joon Yoo1, Jongwon Choi2 1 Image Vision, NAVER Cloud. 2 Department of Advanced Imaging (GSAIM) and Graduate School of AI, Chung-Ang University.
Pseudocode Yes Algorithm 1: Pseudo-code of TVQ-VAE generation" and "Algorithm 2: Pseudo-code of TVQ-VAE training
Open Source Code Yes Official implementation of the proposed TVQ-VAE is available at https: //github.com/clovaai/TVQ-VAE.
Open Datasets Yes We conduct experiments on two datasets: 20 Newsgroups (20NG) (Lang 1995), the New York Timesannotated corpus (NYT) (Sandhaus 2008)... two image datasets: CIFAR10 (Krizhevsky, Hinton et al. 2009) and Celeb A (Liu et al. 2015)
Dataset Splits No The paper mentions 'test set' in the context of evaluation but does not specify explicit training, validation, and test dataset splits with percentages or counts.
Hardware Specification No The paper does not specify any hardware details such as GPU/CPU models, processor types, or memory used for running the experiments.
Software Dependencies No The paper mentions tools like 'sentence Bert' and 'Word2Vec' but does not provide specific version numbers for these or any other software dependencies.
Experiment Setup Yes For TVQ-VAE, we set the embedding number and expansion k to 300 and 5.