Scalable Activation Function Employment in Higher Order Neural Networks in Tasks of Supervised Learning

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F68407700%3A21220%2F23%3A00366751" target="_blank" >RIV/68407700:21220/23:00366751 - isvavai.cz</a>

Výsledek na webu

<a href="http://ysesm2023.itam.cas.cz/index.php" target="_blank" >http://ysesm2023.itam.cas.cz/index.php</a>

DOI - Digital Object Identifier

—

Jazyk výsledku

angličtina

Název v původním jazyce

Scalable Activation Function Employment in Higher Order Neural Networks in Tasks of Supervised Learning

Popis výsledku v původním jazyce

Present contribution deals with methods of supervised learning. A novel approach - scalable activation function and its influence to speed of convergence is studied. Furthermore, unlike classical deep learning approaches, presented neural networks architectures are built with higher order neural units that enable to capture higher levels of non-linearity in tasks. Basics of higher order neural networks and learnable activation function are introduced. The applicability of aforementioned approach is tested and proved on various well-known task of supervised learning, e.g., unknown non-linear functions introduced by Klassen, Gupta et al. and well-known logic function - XOR problem. It turned out that presented method outperformed standard deep learning methods in terms of convergence speed.

Název v anglickém jazyce

Scalable Activation Function Employment in Higher Order Neural Networks in Tasks of Supervised Learning

Popis výsledku anglicky

Present contribution deals with methods of supervised learning. A novel approach - scalable activation function and its influence to speed of convergence is studied. Furthermore, unlike classical deep learning approaches, presented neural networks architectures are built with higher order neural units that enable to capture higher levels of non-linearity in tasks. Basics of higher order neural networks and learnable activation function are introduced. The applicability of aforementioned approach is tested and proved on various well-known task of supervised learning, e.g., unknown non-linear functions introduced by Klassen, Gupta et al. and well-known logic function - XOR problem. It turned out that presented method outperformed standard deep learning methods in terms of convergence speed.

Druh

O - Ostatní výsledky

CEP obor

—

OECD FORD obor

10102 - Applied mathematics

Projekt

<a href="/cs/project/EF16_019%2F0000826" target="_blank" >EF16_019/0000826: Centrum pokročilých leteckých technologií</a><br>

Návaznosti

P - Projekt vyzkumu a vyvoje financovany z verejnych zdroju (s odkazem do CEP)<br>S - Specificky vyzkum na vysokych skolach

Rok uplatnění

2023

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů