The Role of Dimensionality Reduction in Classification
Authors: Weiran Wang, Miguel Carreira-Perpinan
AAAI 2014 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We apply it to a RBF mapping with a linear SVM. This alternates steps where we train the RBF mapping and a linear SVM as usual regression and classification, respectively, with a closedform step that coordinates both. The resulting nonlinear low-dimensional classifier achieves classification errors competitive with the state-of-the-art but is fast at training and testing, and allows the user to trade off runtime for classification accuracy easily. We then study the role of nonlinear DR in linear classification, and the interplay between the DR mapping, the number of latent dimensions and the number of classes. |
| Researcher Affiliation | Academia | Weiran Wang and Miguel A. Carreira-Perpi n an Electrical Engineering and Computer Science, School of Engineering, University of California, Merced |
| Pseudocode | No | The paper describes the algorithm steps (g-step, F-step, Z-step) in paragraph form, but does not present them as a structured pseudocode block or algorithm listing. |
| Open Source Code | No | The paper mentions using LIBSVM (Chang and Lin 2011) and the Matlab Parallel Processing Toolbox, which are third-party tools. There is no explicit statement or link indicating that the authors' own code for the described methodology is open-source or publicly available. |
| Open Datasets | Yes | We perform binary classification on the two topics comp.sys.ibm.pc.hardware and comp.sys.mac.hardware from the 20 newsgroup dataset using the TFIDF features. We now consider the problem of classifying the 10 digit classes of MNIST. |
| Dataset Splits | Yes | For evaluation, we create 10 different 80/20 splits of the 1 162 training items into training and validation set. We randomly sample 10 000 images for training and 10 000 for validation. |
| Hardware Specification | No | The paper mentions running a parallel version of the algorithm using the Matlab Parallel Processing Toolbox, but it does not specify any details about the CPU, GPU, memory, or specific hardware models used for the experiments. |
| Software Dependencies | No | The paper mentions using LIBSVM and the Matlab Parallel Processing Toolbox. However, it does not provide specific version numbers for these tools or any other software dependencies, which are necessary for full reproducibility. |
| Experiment Setup | Yes | We use a initial value of µ = 2 for the quadratic penalty parameter and increase it times 1.5 when the alternating scheme converges for fixed µ. We use early stopping, by exiting the iteration when the error in a validation set does not change or goes up. The hyperparameters are the usual ones: M, σ, λ for the RBF mapping F, and C for the SVM g, and can be determined by cross-validation. |