Understanding Distributed Representations of Concepts in Deep Neural Networks without Supervision

Authors: Wonjoon Chang, Dahee Kwon, Jaesik Choi

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

Reproducibility Variable Result LLM Response
Research Type Experimental In this section, we present the qualitative and quantitative evaluation results of our proposed method as well as various use cases. Our experiments are conducted on the Mini Image Net (Vinyals et al. 2016), Flowers Recognition (denoted by Flowers), Oxford pet, Broden (Bau et al. 2017), Imagenet-X (Idrissi et al. 2022) datasets, using VGG19 (Simonyan and Zisserman 2014), Res Net50 (He et al. 2016), and Mobile Net V2 (Sandler et al. 2018) models.
Researcher Affiliation Collaboration Wonjoon Chang1 *, Dahee Kwon1 *, Jaesik Choi1, 2 1 Korea Advanced Institute of Science and Technology 2 INEEJI
Pseudocode Yes Algorithm 1: Finding a Relaxed Decision Region
Open Source Code No The paper does not provide any information about open-source code availability or links to code repositories.
Open Datasets Yes Our experiments are conducted on the Mini Image Net (Vinyals et al. 2016), Flowers Recognition (denoted by Flowers), Oxford pet, Broden (Bau et al. 2017), Imagenet-X (Idrissi et al. 2022) datasets
Dataset Splits No The paper mentions "training data" and refers to "validation" in the quantitative evaluation but does not provide specific percentages, counts, or explicit methodologies for dataset splits for training, validation, or testing. It mentions: "We empirically check that the RDR works effectively with the parameters k [5, 10] and t [9, 15] in the penultimate convolutional block of the models in our experiments." which is a range and not a concrete value.
Hardware Specification No The paper does not explicitly describe the specific hardware (e.g., GPU/CPU models, memory) used for running the experiments.
Software Dependencies No The paper does not provide specific software names with version numbers that are critical dependencies for reproducibility.
Experiment Setup No The paper states, "Detailed settings of each experiment are provided in Appendix." and mentions