Respect the model: Fine-grained and Robust Explanation with Sharing Ratio Decomposition

Authors: Sangyu Han, Yearim Kim, Nojun Kwak

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

Reproducibility Variable Result LLM Response
Research Type Experimental In this section, we conducted a comprehensive comparative analysis involving our proposed method, SRD, and several state-of-the-art methods: Saliency (Simonyan et al., 2014), Guided Backprop (Springenberg et al., 2015), Grad Input (Ancona et al., 2018), Inte Grad (Sundararajan et al., 2017), LRPz+ (Montavon et al., 2017), Smoothgrad (Smilkov et al., 2017), Fullgrad (Srinivas & Fleuret, 2019), Grad CAM (Selvaraju et al., 2017), Grad CAM++ (Chattopadhay et al., 2018), Score CAM (Wang et al., 2020), Ablation CAM (Ramaswamy et al., 2020), XGrad CAM (Fu et al., 2020), and Layer CAM (Jiang et al., 2021).
Researcher Affiliation Academia Sangyu Han , Yearim Kim , Nojun Kwak Seoul National University Seoul, 08826, Korea {acoexist96,yerim1656,nojunk}@snu.ac.kr
Pseudocode Yes Algorithm 1 APOP process in Py Torch pseudocode
Open Source Code No The paper does not provide concrete access to source code (specific repository link, explicit code release statement, or code in supplementary materials) for the methodology described in this paper.
Open Datasets Yes All evaluations were carried out on the Image Net-S50 dataset (Gao et al., 2022), which contains 752 samples along with object segmentation masks.
Dataset Splits Yes All of our APOP experiments were conducted on the Image Net validation dataset.
Hardware Specification No The paper does not provide specific hardware details (exact GPU/CPU models, processor types with speeds, memory amounts, or detailed computer specifications) used for running its experiments.
Software Dependencies No The paper does not provide specific ancillary software details (e.g., library or solver names with version numbers like Python 3.8, CPLEX 12.4) needed to replicate the experiment.
Experiment Setup Yes The dimensions of the resulting saliency maps were as follows: (7, 7) for low-resolution, (28, 28) for high-resolution, and (224, 224) for input-scale.