Small-Variance Asymptotics for Dirichlet Process Mixtures of SVMs

Authors: Yining Wang, Jun Zhu

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

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results show that M2DPM runs much faster than similar algorithms without sacrificing prediction accuracies. Experimental results on both synthetic and real datasets demonstrate the efficiency and effectiveness of the M2DPM algorithm compared to other competitors.
Researcher Affiliation Academia Yining Wang The Institute for Theoretical Computer Science Institute for Interdisciplinary Information Sciences Tsinghua University, Beijing, China ynwang.yining@gmail.com Jun Zhu Department of Computer Science and Technology TNList Lab, State Key Lab of Intell. Tech. & Sys. Tsinghua University, Beijing, China dcszj@mail.tsinghua.edu.cn
Pseudocode Yes Algorithm 1 The M2DPM algorithm
Open Source Code No The paper does not provide any explicit statement or link regarding the public availability of its source code.
Open Datasets Yes The first real dataset was created in (Ding and Dubchak 2001) for protein fold classification. The second real dataset is described in (Little et al. 2009) for detecting Parkinson s disease.
Dataset Splits Yes In each dataset, we randomly pick 80% instances for training and use the rest 20% for testing. Hyper-parameters are determined by 5-fold cross-validation on training data. We select the hyper-parameters of M2DPM by a 5-fold cross-validation on the training set. We follow the previous setup (i.e., performing 5-fold cross-validation) and compare with the results reported in (Shahbaba and Neal 2009) in Table 4.
Hardware Specification No The paper does not provide specific hardware details (e.g., CPU/GPU models, memory amounts) used for running its experiments.
Software Dependencies No The paper does not provide specific software dependency details with version numbers (e.g., library or solver names with versions).
Experiment Setup Yes Hyper-parameters are determined by 5-fold cross-validation on training data. The algorithm terminates when the relative change of the loss function is less than ε = 10 3. The results are reported using hyperparameters λ = 150, s = 0.01, ν = 1 and c = 2.5, which were selected by a 5-fold cross-validation performed on the data set.