Nature-inspired algorithms for hyperparameter optimization : magistrsko delo

Glojnarić, Filip

| | SLO | ENG | Cookies and privacy

Bigger font | Smaller font

First page > Show document

Show document

Title:	Nature-inspired algorithms for hyperparameter optimization : magistrsko delo
Authors:	ID Glojnarić, Filip (Author) ID Fister, Iztok (Mentor) More about this mentor... ID Brezočnik, Lucija (Comentor)
Files:	MAG_Glojnaric_Filip_2019.pdf (969,13 KB) MD5: DF861200FD42CBBB12C661F032F77885 PID: 20.500.12556/dkum/01924d84-d3d2-495b-b8b7-90e45fe03dd1
Language:	English
Work type:	Master's thesis/paper
Typology:	2.09 - Master's Thesis
Organization:	FERI - Faculty of Electrical Engineering and Computer Science
Abstract:	This master thesis is focusing on the utilization of nature-inspired algorithms for hyperparameter optimization, how they work and how to use them. We present some existing methods for hyperparameter optimization as well as propose a novel method that is based on six different nature-inspired algorithms: Firefly algorithm, Grey Wolf Optimizer, Particle Swarm Optimization, Genetic algorithm, Differential Evolution, and Hybrid Bat algorithm. We also show the optimization results (set of hyperparameters) for each algorithm and we present the plots of the accuracy for each combination and handpicked one. In discussion of the results, we provide the answers on our research questions as well as propose ideas for future work.
Keywords:	artificial intelligence, artificial neural networks, machine learning, nature-inspired algorithms, evolutionary algorithms
Place of publishing:	Maribor
Place of performance:	Maribor
Publisher:	[F. Glojnarić]
Year of publishing:	2019
Number of pages:	XII, 59 f.
PID:	20.500.12556/DKUM-75438
UDC:	004.8.021(043.2)
COBISS.SI-ID:	22917398
NUK URN:	URN:SI:UM:DK:NGYSOWKS
Publication date in DKUM:	09.12.2019
Views:	2232
Downloads:	122
Metadata:
Categories:	KTFMB - FERI
:	GLOJNARIĆ, Filip, 2019, Nature-inspired algorithms for hyperparameter optimization : magistrsko delo [online]. Master’s thesis. Maribor : F. Glojnarić. [Accessed 30 March 2025]. Retrieved from: https://dk.um.si/IzpisGradiva.php?lang=eng&id=75438 Copy citation

Average score:	0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 (0 votes)
Your score:	Voting is allowed only for logged in users.
Share:

Similar works from our repository:

Filling the gaps in diagnostics of Pepino mosaic virus and Potato spindle tuber viroid in water and tomato seeds and leaves
Rapid loop-mediated isothermal amplification assays for grapevine yellows phytoplasmas on crude leaf-vein homogenate has the same performance as qPCR

Similar works from other repositories:

A loop-mediated isothermal amplification (LAMP) assay for the rapid detection of Enterococcus spp. in water
Vzpostavitev metode za ugotavljanje prisotnosti genov sitA, traT in ompT pri sevih Escherichia coli iz perutnine
Razvoj metode z zanko posredovanega izotermalnega pomnoževanja za določanje gliv rodu Verticillium
Strategies for biofilm inhibition and virulence attenuation of foodborne pathogen-Escherichia coli O157:H7
Differences in recipient ability of uropathogenic Escherichia coli strains in relation with their pathogenic potential

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:	14.11.2019

Secondary language

Language:	Slovenian
Title:	Algoritmi po vzorih iz narave za optimizacijo hiperparametrov
Abstract:	V magistrskem delu smo se osredotočili na uporabo algoritmov po vzorih iz narave za optimizacijo hiperparametrov. Predstavili smo strojno učenje, optimizacijske metode in podrobneje šest algoritmov po vzorih iz narave. Ti so algoritem kresnic, algoritem sivega volka, algoritem roja delcev, genetski algoritem, diferencialna evolucija, algoritem po vzoru obnašanja netopirjev in njegova hibridna različica. Pregled literature je pokazal, da so omenjeni algoritmi zelo uporabni pri optimizaciji hiperparametrov, zato nas je analiza spodbudila k predlaganju lastne metode. Sicer so osnovni pristopi, ki se običajno uporabljajo za optimizacijo hiperparametrov, ročno iskanje, Grid search in Bayesove metode. Naša metoda temelji na šestih algoritmih po vzorih iz narave. Z njo skušamo poiskati optimalen nabor hiperparametrov za umetno nevronsko mrežo. Funkcija uspešnosti temelji na točnosti klasifikacije. Kandidatne rešitve so predstavljene kot vektorji realnih števil. Struktura umetne nevronske mreže je z enim skritim slojem, druge nastavitve pa so privzete (razen tistih, ki jih optimiziramo). Program je bil napisan v programskem jeziku Python. Programske knjižnice, ki smo jih uporabili, so Scikit-learn in NiaPy. Eksperiment smo izvedli na treh različnih množicah podatkov in ga zagnali na prenosnem računalniku Dell s 2,67 GHz procesorjem in 4,00 GB pomnilnikom. Operacijski sistem je Windows 10 Education. Za vsak algoritem smo nastavili zaključni pogoj na 10000 ovrednotenj funkcije uspešnosti in velikost populacije na 100 posameznikov. Rezultati nakazujejo, da je uporaba naše metode za HPO učinkovita. Nad dvema množicama podatkov je hibridna različica po vzoru obnašanja netopirjev dosegla najboljši rezultat, nad eno množico podatkov pa je najboljši rezultat dosegel algoritem kresnic. V prihodnosti želimo našo rešitev kot storitev objaviti na spletu. Dodatno bi lahko s tehnikami adaptacije in hibridizacije izboljšali osnovne algoritme po vzorih iz narave, ki smo jih uporabili v tem magistrskem delu.
Keywords:	umetna inteligenca, umetne nevronske mreže, strojno učenje, algoritmi po vzorih iz narave, evolucijski algoritmi

Comments

Leave comment

You must log in to leave a comment.

Comments (0)

0 - 0 / 0

There are no comments!

Back