Prenos znanja med modeli nevronskih mrež z metodo destilacije : diplomsko delo

Marušič, Matic

| | SLO | ENG | Cookies and privacy

Bigger font | Smaller font

First page > Show document

Show document

Title:	Prenos znanja med modeli nevronskih mrež z metodo destilacije : diplomsko delo
Authors:	ID Marušič, Matic (Author) ID Strnad, Damjan (Mentor) More about this mentor... ID Borovič, Mladen (Comentor)
Files:	UN_Marusic_Matic_2022.pdf (2,54 MB) MD5: 36615E8D40E135BFC2316EB8EC829E1E
Language:	Slovenian
Work type:	Bachelor thesis/paper
Typology:	2.11 - Undergraduate Thesis
Organization:	FERI - Faculty of Electrical Engineering and Computer Science
Abstract:	S pojavom vedno večjih zbirk podatkov se je pojavila tudi potreba po vedno večjih modelih strojnega učenja za učenje in napovedovanje na teh zbirkah, kot so nevronske mreže. Posledica tega je nezmožnost uporabe na napravah z omejenimi prostorskimi in računskimi viri, npr. na pametnih mobilnih napravah, pametnih urah in kamerah. Potencialna rešitev je prenos znanja, kjer znanje večjega modela nevronske mreže destiliramo v pomanjšan model nevronske mreže, ki se lahko potem uporablja na robnih napravah. V diplomskem delu smo preizkusili koncept destilacije na klasifikacijskih problemih, ga prenesli na regresijske probleme ter analizirali učinkovitost z uporabo klasičnih metrik uspešnosti.
Keywords:	strojno učenje, destilacija znanja, prenos znanja, nevronska mreža
Place of publishing:	Maribor
Place of performance:	Maribor
Publisher:	[M. Marušič]
Year of publishing:	2022
Number of pages:	1 spletni vir (1 datoteka PDF (IX, 35 f.))
PID:	20.500.12556/DKUM-82955
UDC:	004.85.032.26(043.2)
COBISS.SI-ID:	138725891
Publication date in DKUM:	24.10.2022
Views:	785
Downloads:	58
Metadata:
Categories:	KTFMB - FERI
:	Copy citation

Average score:	(0 votes)
Your score:	Voting is allowed only for logged in users.
Share:

Hover the mouse pointer over a document title to show the abstract or click on the title to get all document metadata.

Licences

License:	CC BY-NC-ND 4.0, Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International

Link:	http://creativecommons.org/licenses/by-nc-nd/4.0/
Description:	The most restrictive Creative Commons license. This only allows people to download and share the work for no commercial gain and for no other purposes.
Licensing start date:	12.09.2022

Secondary language

Language:	English
Title:	Knowledge transfer between neural network models using the distillation method
Abstract:	With the emergence of ever-growing data sets, the need for larger machine learning models, such as neural networks, has also arisen. Consequently, they cannot be used on devices with limited space and computing resources, such as smart mobile devices, smart watches and cameras. A potential solution is knowledge transfer, where the knowledge of a larger neural network model is distilled into a downscaled neural network model that can then be used on edge devices. In the thesis, we tested the concept of distillation on classification problems, transferred it to regression problems, and analyzed the efficiency using classic performance metrics.
Keywords:	machine learning, knowledge distillation, knowledge transfer, neural network

Comments

Leave comment

You must log in to leave a comment.

Comments (0)

0 - 0 / 0

There are no comments!

Back