EURO-Online login
- New to EURO? Create an account
- I forgot my username and/or my password.
- Help with cookies
(important for IE8 users)
4236. Conic Functions: Theory and Applications in Machine Learning
Invited abstract in session WC-27: Machine Learning and Ensemble Learning with optimization methods, stream Mathematical Optimization for XAI.
Wednesday, 12:30-14:00Room: 047 (building: 208)
Authors (first author is the speaker)
| 1. | Gurkan Ozturk
|
| Industrial Engineering, Eskisehir Technical University | |
| 2. | Emre Çimen
|
| Industrial Engineering, Eskisehir Technical Universty | |
| 3. | Zehra Kamisli Ozturk
|
| Industrial Engineering Department, Eskisehir Technical University |
Abstract
Polyhedral conic functions, introduced by Gasimov and Öztürk, are defined with graphs depicting convex polyhedra as their level sets. They have become vital in machine learning, particularly for addressing classification problems. Derived by augmenting a hyperplane equation with the l1 norm, a polyhedral conic function relocates within the space through the vertex point.
The PCF (Polyhedral Conic Function) algorithm efficiently separates two sets of points in multidimensional spaces by solving a linear programming subproblem. It computes parameters and yields a function partitioning the space into two regions, thereby separating points effectively.
PCFs find utility in classification by formulating error functions to minimize discrepancies between sets. They have applications in image processing and psychiatry, aiding tasks like object recognition and disease identification.
Various PCF-based algorithms, like clustering-based PCF and Incremental Conic Functions, use clustering approaches to find vertex points. These extend traditional PCF capabilities by incorporating clustering techniques.
Mathematical programming and optimization play a crucial role in PCF algorithms, providing a framework for fine-tuning performance. Their termination in finite steps underscores their efficiency, making them appealing in real-world applications.
In conclusion, PCFs offer a versatile approach to classification problems, supported by mathematical principles.
Keywords
- Artificial Intelligence
- Machine Learning
- Mathematical Programming
Status: accepted
Back to the list of papers