University of North Texas
An Analysis, Software Implementation, and Applications of Discrete-Time K-Winners-Take-All Neural Network
Abstract
dc:descriptionThe K-Winners-Take-All Neural Networks (KWTA NNs) identify the K largest or smallest among N inputs, where 1 ≤ K < N, and both K and N are any finite integers. These networks produce positive outputs corresponding to the K largest inputs, and negative (or zero) outputs for the remaining N − K inputs. KWTA NNs are used in numerous applications such as pattern classification, pattern recognition, parallel sorting, parallel filtering, adversarial defense, dynamic tracking of robot systems, and telecommunications. In the scope of this thesis, the discrete-time KWTA NN is analyzed. The existence and uniqueness of the steady states of the discrete-time state variable trajectories of the network are shown. It is proved that these trajectories are globally stable and globally convergent to their steady states. It is also demonstrated that the network reaches the steady states in a finite number of time steps. Furthermore, the network’s behavior in the presence of perturbations in its nonlinearity is analyzed. The KWTA NN is compared with other competitive analogs and employed in parallel sorting and parallel rank-order filtering. Computer simulations that confirm the analytical results and illustrate the network’s performance are provided in Matlab, Python, and C++.
Degree
thesis:*- Grantor dc:publisher
- University of North Texas
- Year dc:date
- 2025
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Vemulapalli, Bavya Sri 2002-
- Contributors dc:contributor
-
- Tunc, Cihan
- Tymoshchuk, Pavlo
- Fu, Song
- Othmane, Lotfi Ben
Subjects
dc:subject × 4Rights
dc:rights- Statement dc:rights
-
- Public
- Vemulapalli, Bavya Sri 2002-
- Copyright
- Copyright is held by the author, unless otherwise noted. All rights Reserved.
- Language dc:language
- English
Identifiers
dc:identifier.*- Identifier
-
local-cont-no: submission_4565
ark: ark:/67531/metadc2566445