Institutional Repository of Vilnius University
Kapsulinių dirbtinių neuroninių tinklų atsparumas geometrinėms transformacijoms /
Abstract
dc:descriptionAdvancing modern technologies are expanding the research possibilities of artificial neuron networks. Because of this, new network architectural improvements are introduced more and more frequently. In this work, newest artificial neuron network architecture – Capsule Network is being studied, which demonstrates a different view of processing data. During the study CapsNet steadiness was experimentally tested by fooling it with geometrical MNIST test data transformations: rotation, translation, scaling and perspective transformations. Experimental results showed that CapsNet is more resistant to misleading than Convolutional Neural Network, however, the difference was not significant enough to prove that CapsNet is vastly superior to CNN. Using neuron groups – capsules in artificial neuron network is a promising idea, but it still requires a considerable improvement.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2018
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Chščanovič, Tomaš,
- Contributors dc:contributor
-
- Valaitis, Vytautas
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD29548649&prefLang=en_US