CoT: Cooperative Training for Generative Modeling of Discrete Data

Authors: Sidi Lu, Lantao Yu, Siyuan Feng, Yaoming Zhu, Weinan Zhang

ICML 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In the experiments, compared to existing state-of-the-art methods, Co T shows superior or at least competitive performance on sample quality, diversity, as well as training stability.
Researcher Affiliation Academia 1APEX Lab, Shanghai Jiao Tong University, Shanghai, China 2Stanford University, California, USA.
Pseudocode Yes Algorithm 1 Cooperative Training
Open Source Code Yes Code for repeatable experiments of this subsection is provided in supplementary materials.
Open Datasets Yes Following the experiment proposed in Leak GAN (Guo et al., 2017), we choose EMNLP 2017 WMT News Section as our dataset, with maximal sentence length limited to 51.
Dataset Splits No The paper mentions evaluating on "test data separated from the original dataset" but does not specify the explicit percentages or sample counts for training, validation, and test splits, nor does it refer to predefined standard splits with citations that include this information. No explicit validation set is described.
Hardware Specification No The paper does not provide specific details about the hardware used for running the experiments, such as GPU models, CPU types, or memory amounts. It does not mention any cloud or cluster specifications.
Software Dependencies No The paper mentions using "Adam (Kingma & Ba, 2014) as the optimizer," but it does not specify version numbers for any software libraries, programming languages (e.g., Python 3.x), or specific frameworks (e.g., PyTorch 1.x) that would be needed for reproducibility.
Experiment Setup Yes We use Adam (Kingma & Ba, 2014) as the optimizer, with hyper-parameter settings of α = 1e-4, β1 = 0.5, β2 = 0.9. The network architecture for fω is shown in Table 4.