Anomaly Ranking as Supervised Bipartite Ranking

Authors: Stephan Clémençon, Sylvain Robbiano

ICML 2014 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Numerical experiments based on a variety of bipartite ranking algorithms well-documented in the literature are displayed in order to illustrate the relevance of our approach.
Researcher Affiliation Academia LTCI UMR Telecom Paris Tech/CNRS No. 5141, 46 rue Barrault, 75634 Paris Cedex, France CIMFAV-Facultad de Ingeniera, Universidad de Valparaso, Valparaso, Chile
Pseudocode No No pseudocode or clearly labeled algorithm blocks were found in the paper.
Open Source Code Yes Precisely, we implemented the modification of the TREERANK procedure based on locally weighted versions of the CART method (with axis parallel splits) described at length in subsection 4.2, using the package for R statistical software (see http://www.rproject.org), available at http://treerank.sourceforge.net (with parameters: minsplit = 1, maxdepth = 4, in the LEAFRANK), see (Baskiotis et al., 2009).
Open Datasets Yes We also used a benchmark dataset in anomaly detection (computer network intrusion detection namely) proposed as a challenge for intrusion detection in the CMDC2013, see http://www.csmining.org/cdmc2013/ and (Song, 2013).
Dataset Splits Yes In the following experiment, an estimate of the area under the MV-curve (AMV in short) is computed over 5 replications of a 5-fold cross validation as well as the overall standard deviation (denoted by σ).
Hardware Specification No No specific hardware details (such as GPU/CPU models, memory, or cloud instance types) used for running experiments were mentioned in the paper.
Software Dependencies No The paper mentions software like 'R statistical software', 'SVM-light implementation', 'Rank RLS method', and 'R-package Kernlab' but does not provide specific version numbers for these dependencies.
Experiment Setup Yes Precisely, we implemented the modification of the TREERANK procedure... (with parameters: minsplit = 1, maxdepth = 4, in the LEAFRANK)... Rank Boost (aggregating 30 stumps)... SVMRank (with linear and Gaussian kernels with cross-validated parameters)... Rank RLS method... with linear kernel ( bias = 1 ) and with Gaussian kernel (γ = 0.01)... R-package Kernlab with gaussian kernel, with parameters chosen automatically by cross-validation.