ALTBI: Constructing Improved Outlier Detection Models via Optimization of Inlier-Memorization Effect

Authors: Seoyoung Cho, Jaesung Hwang, Kwan-Young Bak, Dongha Kim

AAAI 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We provide extensive experimental results to demonstrate that ALTBI achieves state-ofthe-art performance in identifying outliers compared to other recent methods, even with lower computation costs. Additionally, we show that our method yields robust performances when combined with privacy-preserving algorithms.
Researcher Affiliation Collaboration 1Department of Statistics, Sungshin Women s University 2SK Telecom 3Data Science Center, Sungshin Women s University EMAIL, EMAIL
Pseudocode Yes The pseudo algorithm of ALTBI is presented in Algorithm 1.
Open Source Code No The paper states, "We acknowledge that we implemented ALTBI and ODIM ourselves," but it does not provide an explicit statement of code release or a link to a code repository for the methodology described in the paper.
Open Datasets Yes We analyze all 57 outlier detection benchmark datasets from ADBench (Han et al. 2022), including tabular, image, and text data.
Dataset Splits No The paper mentions evaluating AUC values of the training data and running experiments for three trials with random parameter initializations, but it does not specify explicit training/test/validation dataset splits (e.g., percentages, sample counts, or references to predefined splits) needed for reproduction.
Hardware Specification Yes We use the Pytorch framework to run our algorithm using a single NVIDIA TITAN XP GPU.
Software Dependencies No The paper mentions using the "Pytorch framework" and "Adam" optimizer, but it does not provide specific version numbers for these software components or any other key libraries.
Experiment Setup Yes For the optimizer, we use Adam (Kingma and Ba 2014) with a learning rate of 1e 3. Throughout our experimental analysis, we fix the hyperparameters, necessary for our proposed method (n0, γ, ρ, T0, T1, T2) to (128,1.03,0.92,10,60,80), unless stated otherwise.