Directional Label Rectification in Adaptive Graph
Authors: Xiaoqian Wang, Hao Huang
AAAI 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We conduct extensive experiments on both synthetic and real world datasets and illustrate the advantage of our model in both effectiveness and efficiency. |
| Researcher Affiliation | Collaboration | Xiaoqian Wang,1 Hao Huang2 1Department of Electrical and Computer Engineering, University of Pittsburgh, Pittsburgh, PA, USA, 15261 2GE Global Research, San Ramon, CA, USA, 94583 |
| Pseudocode | Yes | Algorithm 1: Optimization Algorithm for DRAG |
| Open Source Code | No | The paper does not provide any concrete access to source code for the described methodology. No links to repositories or explicit statements about code availability are present. |
| Open Datasets | Yes | The synthetic dataset 1 is constructed on the basis of Yahoo Benchmark Dataset for Time Series Anomaly Detection (Laptev, Amizadeh, and Youssef Mar 25 2015). 1https://goo.gl/16l M7h |
| Dataset Splits | No | The paper describes how instances Xa and Xb are defined and used in the problem setting, and discusses evaluation metrics (F1-score in linear SVM classification), but it does not provide specific percentages or counts for training, validation, or test dataset splits. |
| Hardware Specification | Yes | All experiments are conducted on a standard laptop (Quadcore Intel i7 CPU@2.7 GHz). |
| Software Dependencies | No | The paper mentions 'linear kernel for all SVM models (implementation is based on libsvm)' but does not provide specific version numbers for libsvm or any other software dependencies. |
| Experiment Setup | Yes | Note that all hyper-parameters of the comparing methods are set according to the reported best or tuned to get the best performance. Specifically, 1) for GMM models, we set the number of components as 6; 2) for Coselect, we set α = 0.1, β = 0.01, λ = 0.3, and construct the linking graph R with the following strategy: if xi and xj belong to the same event in Xa then R(i, j) = 1, otherwise zero; 3) for DRAG, the relative magnitude relation of entries in Ya and W is relatively stable w.r.t. parameters, so in Algorithm 1 we set the following by default: α = m/k, β = 1/b, μ = s/2a and γ = 1. |