Institutional Repository [SANDBOX]
Technical University of Crete
EN  |  EL

Search

Browse

My Space

Distributed training of recurrent neural networks by FGM protocol

Balampanis Ilias

Full record


URI: http://purl.tuc.gr/dl/dias/E93A906F-856D-4D07-8B71-93114448146E
Year 2020
Type of Item Diploma Work
License
Details
Bibliographic Citation Ilias Balampanis, "Distributed training of recurrent neural networks by FGM protocol", Diploma Work, School of Electrical and Computer Engineering, Technical University of Crete, Chania, Greece, 2020 https://doi.org/10.26233/heallink.tuc.86833
Appears in Collections

Summary

Artificial Neural Networks are appealing because they learn by example and are strongly supported by statistical and optimization theories. The usage of recurrent neural networks as identifiers and predictors in nonlinear dynamic systems has increased significantly. They can present a wide range of dynamics, due to feedback and are also flexible nonlinear maps. Based on this, there is a need for distributed training on these networks, because of the enormous datasets. One of the most known protocols for distributed training is the Geometric Monitoring protocol. Our conviction is that this is a very expensive protocol regarding the communication of nodes. Recently, the Functional Geometric Protocol has tested training on Convolutional Neural Networks and has had encouraging results. The goal of this work is to test and compare these two protocols on Recurrent Neural Networks.

Available Files

Services

Statistics