Learning Dynamics of Attention: Human Prior for Interpretable Machine Reasoning

Authors: Wonjae Kim, Yoonho Lee

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

Reproducibility Variable Result LLM Response
Research Type Experimental Our experiments reveal that applying DAFT yields similar performance to the original model while using fewer reasoning steps, showing that it implicitly learns to skip unnecessary steps. We conducted our experiments on the CLEVR3 [Johnson et al., 2017a] and GQA4 [Hudson and Manning, 2019] datasets.
Researcher Affiliation Industry Wonjae Kim Kakao Corporation Pangyo, Republic of Korea dandelin.kim@kakaocorp.com Yoonho Lee Kakao Corporation Pangyo, Republic of Korea eddy.l@kakaocorp.com
Pseudocode Yes Algorithm 1 Memory Update Procedure of MAC; Algorithm 2 Memory Update Procedure of DAFT MAC
Open Source Code Yes Our code is available at https://github.com/kakao/DAFT.
Open Datasets Yes We conducted our experiments on the CLEVR3 [Johnson et al., 2017a] and GQA4 [Hudson and Manning, 2019] datasets.
Dataset Splits Yes CLEVR has 700K questions for training and 150K questions for validation and test split.
Hardware Specification No The paper does not explicitly describe the specific hardware used (e.g., GPU models, CPU types, memory) for running its experiments.
Software Dependencies No The paper does not provide specific ancillary software details with version numbers (e.g., library or solver names with version numbers) in the provided text.
Experiment Setup No The paper states: 'For a fair comparison, we used the same hyperparameters as the original MAC network [Hudson and Manning, 2018] and closely followed their experimental setup. [...] We list implementation details in Appendix B.' However, it defers the specific details to another paper and an appendix not provided in the main text, thus not explicitly listing concrete hyperparameter values or training configurations within the main body.