|
| |
|
|
Probabilistic Neural Networks for Special Tasks in Electromagnetics
Koudelka, Vlastimil ; Tučková,, Jana (oponent) ; Hartnagel, Hans Ludwig (oponent) ; Raida, Zbyněk (vedoucí práce)
The thesis deals with behavioural modelling techniques capable solving special tasks in electromagnetics which can be formulated as approximation, classification, probability estimation, and combinatorial optimization problems. Concept of the work lies in applying a probabilistic approach to behavioural modelling. Examined methods address two general problems in machine learning and combinatorial optimization: ”bias vs. variance dilemma” and NP computational complexity. The Boltzmann machine is employed to simplify a complex impedance network. The Parzen window is regularized using the Bayesian strategy for obtaining a model selection criterion for probabilistic and general regression neural networks.
|
|
|
Probabilistic Neural Networks for Special Tasks in Electromagnetics
Koudelka, Vlastimil ; Tučková,, Jana (oponent) ; Hartnagel, Hans Ludwig (oponent) ; Raida, Zbyněk (vedoucí práce)
The thesis deals with behavioural modelling techniques capable solving special tasks in electromagnetics which can be formulated as approximation, classification, probability estimation, and combinatorial optimization problems. Concept of the work lies in applying a probabilistic approach to behavioural modelling. Examined methods address two general problems in machine learning and combinatorial optimization: ”bias vs. variance dilemma” and NP computational complexity. The Boltzmann machine is employed to simplify a complex impedance network. The Parzen window is regularized using the Bayesian strategy for obtaining a model selection criterion for probabilistic and general regression neural networks.
|
|
| |
host ::
RSS.