Kernel Mean Shrinkage Estimators

Authors: Krikamol Muandet, Bharath Sriperumbudur, Kenji Fukumizu, Arthur Gretton, Bernhard Schölkopf

JMLR 2016 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In Section 6, we empirically evaluate the proposed shrinkage estimators of kernel mean on both synthetic data and several real-world scenarios including Parzen window classification, density estimation and discriminative learning on distributions. The experimental results demonstrate the benefits of our shrinkage estimators over the standard one.
Researcher Affiliation Academia Krikamol Muandet EMAIL Empirical Inference Department, Max Planck Institute for Intelligent Systems Spemannstraße 38, T ubingen 72076, Germany Bharath Sriperumbudur EMAIL Department of Statistics, Pennsylvania State University University Park, PA 16802, USA Kenji Fukumizu EMAIL The Institute of Statistical Mathematics 10-3 Midoricho, Tachikawa, Tokyo 190-8562 Japan Arthur Gretton EMAIL Gatsby Computational Neuroscience Unit, CSML, University College London Alexandra House, 17 Queen Square, London WC1N 3AR, United Kingdom Bernhard Sch olkopf EMAIL Empirical Inference Department, Max Planck Institute for Intelligent Systems Spemannstraße 38, T ubingen 72076, Germany
Pseudocode No The paper describes methods mathematically and textually but does not include any explicitly labeled pseudocode or algorithm blocks.
Open Source Code No The paper does not contain any explicit statements about releasing source code or provide links to a code repository for the described methodology.
Open Datasets Yes For some of these tasks we employ datasets from the UCI repositories.
Dataset Splits Yes We use 30% of each dataset as a test set and the rest as a training set... All hyper-parameters are chosen by 10-fold cross-validation.
Hardware Specification No The paper does not provide specific details about the hardware (e.g., CPU, GPU models) used for running the experiments.
Software Dependencies No The paper does not specify any software dependencies with version numbers.
Experiment Setup Yes Unless otherwise stated, we set the bandwidth parameter of the Gaussian kernel as σ2 = median xi xj 2 : i, j = 1, . . . , n , i.e., the median heuristic... In this experiment, we only consider the Gaussian RBF kernel whose bandwidth parameter is chosen by cross-validation procedure over a uniform grid σ [0.1, 2]... For each dataset, we set the number of mixture components m to be 10. The model is initialized by running 50 random initializations using the k-means algorithm and returning the best.