Title
Развој кодера таласног облика за потребе неуронских мрежа и обраду сигнала
Creator
Aleksić, Danijela R., 1977-
CONOR:
92608521
Copyright date
2022
Object Links
Select license
Autorstvo-Nekomercijalno-Bez prerade 3.0 Srbija (CC BY-NC-ND 3.0)
License description
Dozvoljavate samo preuzimanje i distribuciju dela, ako/dok se pravilno naznačava ime autora, bez ikakvih promena dela i bez prava komercijalnog korišćenja dela. Ova licenca je najstroža CC licenca. Osnovni opis Licence: http://creativecommons.org/licenses/by-nc-nd/3.0/rs/deed.sr_LATN. Sadržaj ugovora u celini: http://creativecommons.org/licenses/by-nc-nd/3.0/rs/legalcode.sr-Latn
Language
Serbian
Cobiss-ID
Theses Type
Doktorska disertacija
description
Datum odbrane: 17.10.2022.
Other responsibilities
predsednik komisije
Jovanović, Aleksandra
član komisije
Ćirić, Dejan
član komisije
Đurović, Željko
član komisije
Stanimirović, Aleksandar
Academic Expertise
Prirodno-matematičke nauke
Academic Title
-
University
Univerzitet u Nišu
Faculty
Elektronski fakultet
Group
Katedra za telekomunikacije
Alternative title
Development of waveform coders for nеural networks appications and signal processing
Publisher
[D. R. Аleksić]
Format
174 str.
description
Bibliografija: str. 163-169.
description
Telecommunication (Digital processing and signal coding)
Abstract (en)
This doctoral thesis aims to design low-bit scalar quantizers and analyze their application in Neural Networks (NNs) and signal processing. In this thesis, we consider the possibilities and limitations that rest on quantization, as a leading technique for data coding and compression. In particular, we examine the inevitable accuracy loss of signal and data presentation due to quantization in the signal processing area, as well as in many modern solutions, that use quantization. As stated in this thesis, there are a number of qualitative performance indicators, which indicate that appropriate quantizer parameterization can optimize the amount of data transmitted in bits. Quantized Neural Networks (QNNs) is a promising research area, especially important for resource constrained devices. Relying on a plethora of conclusions about scalar quantizers derived for signal processing tasks and taking into account the advantages of scalar quantization, we anticipate that by studying the statistical characteristics of neural network parameters, this thesis will contribute to determining an efficient weights compression solution utilizing new, well-designed scalar quantizers for post-training quantization.
Authors Key words
Skalarna kvantizacija, niskobitni kvantizeri, Laplasova funkcija gustine verovatnoće, odnos signal-šum kvantizacije, neuronske mreže, tačnost neuronskih mreža, kvantovane neuronske mreže, post-trening
Authors Key words
Scalar Qunatization, Low-bit quantizer, Laplacian probability density function, SQNR, Neural Network, Accuracy of Neural Network, Quantized Neual Network, Post-Training
Classification
621.391:004.7(043.3)
Subject
T 121
Type
Tekst
Abstract (en)
This doctoral thesis aims to design low-bit scalar quantizers and analyze their application in Neural Networks (NNs) and signal processing. In this thesis, we consider the possibilities and limitations that rest on quantization, as a leading technique for data coding and compression. In particular, we examine the inevitable accuracy loss of signal and data presentation due to quantization in the signal processing area, as well as in many modern solutions, that use quantization. As stated in this thesis, there are a number of qualitative performance indicators, which indicate that appropriate quantizer parameterization can optimize the amount of data transmitted in bits. Quantized Neural Networks (QNNs) is a promising research area, especially important for resource constrained devices. Relying on a plethora of conclusions about scalar quantizers derived for signal processing tasks and taking into account the advantages of scalar quantization, we anticipate that by studying the statistical characteristics of neural network parameters, this thesis will contribute to determining an efficient weights compression solution utilizing new, well-designed scalar quantizers for post-training quantization.
“Data exchange” service offers individual users metadata transfer in several different formats. Citation formats are offered for transfers in texts as for the transfer into internet pages. Citation formats include permanent links that guarantee access to cited sources. For use are commonly structured metadata schemes : Dublin Core xml and ETUB-MS xml, local adaptation of international ETD-MS scheme intended for use in academic documents.