Rekurento neironu tīklu novērtēšanas sistēma
Loading...
Date
Authors
Advisor
Journal Title
Journal ISSN
Volume Title
Publisher
Latvijas Universitāte
Language
lav
Abstract
Darba “Rekurento neironu tīklu novērtēšanas sistēma” ietvaros tika izstrādāta sistēma, kurā var uztrenēt rekurento neironu tīklu arhitektūras uz dažādiem uzdevumiem, tām automātiski piemeklējot optimālos tīkla parametrus, un pārbaudīt uz tiem sasniegtos rezultātus, ar mērķi objektīvi novērtēt konkrēta rekurentā neironu tīkla efektivitāti. Sistēmu ir paredzēts lietot dziļās mašīnmācīšanās pētniecībā, jaunu rekurento neironu tīklu izstrādē un testēšanā, lai noteiktu vai jaunizveidotie neironu tīkli ir labāki par jau eksistējošajiem uz konkrētiem uzdevumiem.
As a part of the work “Recurrent neural network evaluation system”, a system was developed in which recurrent neural network architectures can be trained on various tasks, automatically selecting the optimal network parameters, and examine the results achieved on them, with the aim to objectively evaluate the efficiency of a specific recurrent neural network. The system is intended to be used in deep learning research, in development and testing of new recurrent neural networks to determine whether the newly created neural networks are better than the ones already existing on specific tasks.
As a part of the work “Recurrent neural network evaluation system”, a system was developed in which recurrent neural network architectures can be trained on various tasks, automatically selecting the optimal network parameters, and examine the results achieved on them, with the aim to objectively evaluate the efficiency of a specific recurrent neural network. The system is intended to be used in deep learning research, in development and testing of new recurrent neural networks to determine whether the newly created neural networks are better than the ones already existing on specific tasks.