2360. Ordinal Cascades of Linear Classifiers
Invited abstract in session WE-7: Simulation, Data & Decision Support, stream Simulation and Quantum Computing.
Wednesday, 16:30-18:00Room: U2-205
Authors (first author is the speaker)
| 1. | Ludwig Maximilian Lausser
|
| Fakultät Informatik, Technische Hochschule Ingolstadt |
Abstract
Ordinal classifier cascades (OCCs) are specialized multi-class classifier systems (MCCS) designed to detect ordinal relations among classes. They can be applied in de novo analyses of large class collections, where they systematically evaluate all potential ordinal (sub-)cascades [1]. For example, they can be utilized to reconstruct developmental processes from gene expression profiles, making them valuable tools in ontogenesis and oncogenesis [2].
However, the performance of ordinal classifier cascades is highly dependent on the selected type of base classifier. Underlying ordinal relations may only be detected if the correct type of base classifier is chosen. In a previous study, linear classifiers outperformed their non-linear counterparts.
In this work, the influence of training algorithms of the base classifiers will be analyzed. In an empirical study, ordinal classifier cascades are coupled with different types of linear classifiers, and the influence on their decision regions is analyzed. The ordinal classifier cascades will be compared based on their generalization performance and their ability to detect ordinal relations among classes.
1. Lausser L, Schäfer LM, Kühlwein SD, Kestler AMR, Kestler HA. Detecting Ordinal Subcascades. Neural Processing Letters, 52(3): 2583-2605, 2020.
2. Lausser L, Schäfer LM, Schirra LR, Szekely R, Schmid F, Kestler HA. Assessing phenotype order in molecular data. Scientific Reports 9(11746), 2019.
Keywords
- Computational Biology
- Machine Learning
- Data Mining
Status: accepted
Back to the list of papers