Kernel Mean Estimation and Stein Effect
Authors: Krikamol Muandet, Kenji Fukumizu, Bharath Sriperumbudur, Arthur Gretton, Bernhard Schoelkopf
ICML 2014 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | 4. Experiments We focus on the comparison between our shrinkage estimators and the standard estimator of the kernel mean using both synthetic datasets and real-world datasets. |
| Researcher Affiliation | Academia | Krikamol Muandet KRIKAMOL@TUEBINGEN.MPG.DE Empirical Inference Department, Max Planck Institute for Intelligent Systems, T ubingen, Germany Kenji Fukumizu FUKUMIZU@ISM.AC.JP The Institute of Statistical Mathematics, Tokyo, Japan Bharath Sriperumbudur BS493@STATSLAB.CAM.AC.UK Statistical Laboratory, University of Cambridge, Cambridge, United Kingdom Arthur Gretton ARTHUR.GRETTON@GMAIL.COM Gatsby Computational Neuroscience Unit, University College London, London, United Kingdom Bernhard Sch olkopf BS@TUEBINGEN.MPG.DE Empirical Inference Department, Max Planck Institute for Intelligent Systems, T ubingen, Germany |
| Pseudocode | No | The paper describes methods mathematically and textually but does not include structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide concrete access to source code or explicitly state that it is being made available. |
| Open Datasets | Yes | For the first two tasks we employ 15 datasets from the UCI repositories. |
| Dataset Splits | Yes | All hyper-parameters are chosen by 10-fold crossvalidation. |
| Hardware Specification | No | The paper does not provide specific details on the hardware used for experiments, such as CPU or GPU models, or memory specifications. |
| Software Dependencies | No | The paper states 'In practice, we use the fminsearch and fminbnd routines of the MATLAB optimization toolbox to find the best shrinkage parameter,' but it does not specify version numbers for MATLAB or the toolbox. |
| Experiment Setup | Yes | For the Gaussian RBF kernel, we set the bandwidth parameter to square-root of the median Euclidean distance between samples in the dataset (i.e., σ2 = median xi xj 2 throughout). Data are generated from a d-dimensional mixture of Gaussians: P mi=1 πi N(θi, Σi) + ε, θij U( 10, 10), Σi W(2 Id, 7), ε N(0, 0.2 Id), where U(a, b) and W(Σ0, df) represent the uniform distribution and Wishart distribution, respectively. We set π = [0.05, 0.3, 0.4, 0.25]. |