|
|
Approximation of functions continuous on compact sets by layered neural networks
Fojtík, Vít ; Hakl, František (advisor) ; Mrázová, Iveta (referee)
Despite abundant research into neural network applications, many areas of the under- lying mathematics remain largely unexplored. The study of neural network expressivity is vital for understanding their capabilities and limitations. However, even for shallow networks this topic is far from solved. We provide an upper bound on the number of neurons of a shallow neural network required to approximate a function continuous on a compact set with given accuracy. Dividing the compact set into small polytopes, we ap- proximate the indicator function of each of them by a neural network and combine these into an approximation of the target function. This method, inspired by a specific proof of the Stone-Weierstrass Theorem, is more general than previous bounds of this character, with regards to approximation of continuous functions. Also, it is purely constructive. 1
|
|
|
Random measurable sets
Fojtík, Vít ; Rataj, Jan (advisor) ; Pawlas, Zbyněk (referee)
The aim of this thesis is to compare two major models of random sets, the well established random closed sets (RACS) and the more recent and more general random measurable sets (RAMS). First, we study the topologies underlying the models, showing they are very different. Thereafter, we introduce RAMS and RACS and reformulate prior findings about their relationship. The main result of this thesis is a characterization of those RAMS that do not induce a corresponding RACS. We conclude by some examples of such RAMS, including a construction of a translation invariant RAMS. 1
|
|
|
Lower Bounds on Boolean Formula Size
Fojtík, Vít ; Hrubeš, Pavel (advisor) ; Savický, Petr (referee)
The aim of this thesis is to study methods of constructing lower bounds on Boolean formula size. We focus mainly on formal complexity measures, gener- alizing the well-known Krapchenko measure to a class of graph measures, which we thereafter study. We also review one of the other main approaches, using ran- dom restrictions of Boolean functions. This approach has yielded the currently largest lower bounds. Finally, we mention a program for finding super-polynomial bounds based on the KRW conjecture. 1
|
guest ::
RSS feed.