Variable Kernel Density Estimation in High-Dimensional Feature Spaces

Authors: Christiaan van der Walt, Etienne Barnard

AAAI 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We compare the performance of this estimator to state-of-the art maximumlikelihood estimators on a number of representative high-dimensional machine learning tasks and show that the newly introduced minimum leave-one-out entropy estimator performs optimally on a number of highdimensional datasets considered.
Researcher Affiliation Academia Christiaan M. van der Walt a,b Etienne Barnard b a Modelling and Digital Science, CSIR, Pretoria, South Africa cvdwalt@csir.co.za b Multilingual Speech Technologies Group, North-West University, Vanderbijlpark, South Africa etienne.barnard@nwu.ac.za
Pseudocode No The paper contains mathematical derivations and equations but no structured pseudocode or algorithm blocks.
Open Source Code No The paper does not provide concrete access to source code for the methodology described, nor does it contain explicit code release statements or links.
Open Datasets Yes We select five RW datasets (with independent train and test sets) from the UCI Machine Learning Repository (Lichman 2013) for the purpose of simulation studies.
Dataset Splits Yes We therefore perform 10-fold cross-validation on each class specific training set to find the optimal number of training iterations for each class conditional density function.
Hardware Specification No The paper does not provide specific hardware details (exact GPU/CPU models, processor types with speeds, memory amounts, or detailed computer specifications) used for running its experiments.
Software Dependencies No The paper does not provide specific ancillary software details (e.g., library or solver names with version numbers) needed to replicate the experiment.
Experiment Setup Yes The optimal kernel bandwidth is obtained with a direct approach where the right-hand side of each bandwidth estimation equation is initialised with the Silverman bandwidth, the left hand side is updated, and the updated bandwidth is then substituted into the right-hand side again. This process is repeated for 10 iterations on each training set.