Improving the k-nearest neighbour rule: using geometrical neighbourhoods and manifold-based metrics

Date

2011-09-01

Author

ALTINCAY, HAKAN

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

140
views

0
downloads

Sample weighting and variations in neighbourhood or data-dependent distance metric definitions are three principal directions considered for improving the k-NN classification technique. Recently, manifold-based distance metrics attracted considerable interest and computationally less demanding approximations have been developed. However, a careful comparison of these alternative approaches is missing. In this study, an extensive comparison is firstly performed for three alternative neighbourhood definitions and four manifold-based distance measures. Then, a novel computationally less demanding feature line-based method is proposed, which exploits geometrical neighbourhoods of test samples for feature line construction. Experimental results have shown that the improvements achieved by the majority of the existing schemes are not considerable. It is also verified that the proposed scheme surpasses other computationally less demanding manifold-based schemes.

Subject Keywords

Control and Systems Engineering, Theoretical Computer Science, Computational Theory and Mathematics, Artificial Intelligence

URI

https://hdl.handle.net/11511/63927

Journal

EXPERT SYSTEMS

DOI

https://doi.org/10.1111/j.1468-0394.2010.00549.x

Collections

Department of Computer Engineering, Article

Suggestions

OpenMETU
Core

Assessment of Transient Stability of Nonlinear Dynamic Systems by the Method of Tangent Hyperplanes and the Method of Tangent Hypersurfaces Eskicioglu, Ahmet M. (ASME International, 1989-9-1) Two direct methods, the method of tangent hyperplanes and the method of tangent hypersurfaces, are applied to an elementary nonlinear dynamic system for transient stability assessment. The former method is based on the approximation of the asymptotic stability boundary by hyperplanes at a certain class of unstable singular points in the state-space, and the latter replaces hyperplanes by hypersurfaces. The applicability and accuracy of both methods are evaluated through a comparison of results.
Quantitative measure of observability for linear stochastic systems Subasi, Yuksel; Demirekler, Mübeccel (Elsevier BV, 2014-06-01) In this study we define a new observability measure for stochastic systems: the mutual information between the state sequence and the corresponding measurement sequence for a given time horizon. Although the definition is given for a general system representation, the paper focuses on the linear time invariant Gaussian case. Some basic analytical results are derived for this special case. The measure is extended to the observability of a subspace of the state space, specifically an individual state and/or t...
Accurate numerical bounds for the spectral points of singular Sturm-Liouville problems over 0 < x < infinity Taşeli, Hasan (Elsevier BV, 2004-03-01) The eigenvalues of singular Sturm-Liouville problems defined over the semi-infinite positive real axis are examined on a truncated interval 0<x<l as functions of the boundary point l. As a basic theoretical result, it is shown that the eigenvalues of the truncated interval problems satisfying Dirichlet and Neumann boundary conditions provide, respectively, upper and lower bounds to the eigenvalues of the original problem. Moreover, the unperturbed system in a perturbation problem, where l remains sufficient...
Developing a computer program for evaluating uncertainty of some typical dimensional measuring and gauging devices Çelebioğlu, Hasan Emrah; Akkök, Metin; Department of Mechanical Engineering (2005) In dimensional measurements, it is required to specify uncertainty in the measurement as the range of possible deviation for the measurement result. In this thesis, a computer program is developed for evaluating uncertainty inmeasurement of commonly used dimensional measuring devices like vernier callipers, micrometers, comparators, and gauge blocks. In evaluation of the uncertainty in measurement, some uncertainty sourceslike temperature difference between the measured part and the instrument, uncertainty ...
Ranking units by target-direction-set value efficiency analysis and mixed integer programming Büyükbaşaran, Tayyar; Köksalan, Murat; Department of Industrial Engineering (2005) In this thesis, two methods are proposed in order to rank units: Target-direction-set value efficiency analysis (TDSVEA) and mixed integer programming (MIP) technique. Besides its ranking ability based on preferences of a decision maker (DM), TDSVEA, which modifies the targeted projection approach of Value Efficiency Analysis (VEA) and Data Envelopment Analysis (DEA), provides important information to analyzer: targets and distances of units from these targets, proposed input allocations in order to project...

Citation Formats

H. ALTINCAY, “Improving the k-nearest neighbour rule: using geometrical neighbourhoods and manifold-based metrics,” EXPERT SYSTEMS, pp. 391–406, 2011, Accessed: 00, 2020. [Online]. Available: https://hdl.handle.net/11511/63927.