Dual-Regularized Multi-View Outlier Detection

Authors: Handong Zhao, Yun Fu

IJCAI 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We evaluate our method on five datasets with different outlier settings. The consistent superior results to other stateof-the-art methods demonstrate the effectiveness of our approach.
Researcher Affiliation Academia Handong Zhao1 and Yun Fu1,2 1 Department of Electrical and Computer Engineering, Northeastern University, Boston, USA, 02115 2 College of Computer and Information Science, Northeastern University, Boston, USA, 02115 {hdzhao,yunfu}@ece.neu.edu
Pseudocode Yes Algorithm 1. Optimization Solution of Problem (2) and Algorithm 2. DMOD for Multi-view Outlier Detection are provided, outlining the steps for the proposed methods.
Open Source Code No The paper does not provide an explicit statement or link indicating that the source code for the described methodology is publicly available.
Open Datasets Yes Among them, four are from UCI Machine Learning Repository1, i.e. iris, breast, ionosphere, and letter. The fifth one Vis Nir is from BUAA database [Di Huang and Wang, 2012].
Dataset Splits No The paper describes how outliers are generated and how performance is evaluated across different outlier ratios. However, it does not specify explicit train/validation/test splits for model development or a cross-validation setup.
Hardware Specification No The paper does not provide specific details about the hardware used for running the experiments (e.g., GPU/CPU models, memory).
Software Dependencies No The paper does not specify any software dependencies with version numbers.
Experiment Setup Yes Three parameters β, γ and K are set to 0.5, 0.1 and 3, respectively. The intrinsic dimension K of iris dataset is 3 since it has three classes. Therefore we make the evaluation in the range of [1, 8]. The average AUCs in three different settings are relatively steady when γ = {10 3, 10 2, 10 1, 100}. In practical, we choose γ = 0.1 as default for all experiments.