| | SLO | ENG | Cookies and privacy

Bigger font | Smaller font

Show document Help

Title:Uporaba nevronskih mrež pri iskanju binarnih sekvenc z nizkimi avtokorelacijami : magistrsko delo
Authors:ID Popič, Jan (Author)
ID Bošković, Borko (Mentor) More about this mentor... New window
ID Brest, Janez (Comentor)
Files:.pdf MAG_Popic_Jan_2023.pdf (2,58 MB)
MD5: 4F8600D06D430A14C53661B5B15A8C66
 
Language:Slovenian
Work type:Master's thesis/paper
Typology:2.09 - Master's Thesis
Organization:FERI - Faculty of Electrical Engineering and Computer Science
Abstract:Uporaba nevronskih mrež je vedno bolj razširjena, tako v vsakdanjem življenju kot na različnih raziskovalnih področjih. Kljub razširjeni uporabi pa obstajajo raziskovalni problemi, kjer uporabna vrednost nevronskih mrež še ni bila preverjena. Eno izmed takšnih področij je iskanje binarnih sekvenc z nizko avtokorelacijo (ang. low-autocorelation binary sequence), pri katerem se iščejo binarna zaporedja različnih dolžin, ki imajo čim manjšo vrednost avtokorelacije. Takšne sekvence se zaradi svojih specifičnih lastnosti uporabljajo pri mnogih raziskovalnih področjih, njihovo iskanje pa predstavlja izjemno zahteven kombinatoričen problem. V našem delu predstavimo dve nevronski mreži, ki služita za usmerjanje iskalnega algoritma samoizogibnega sprehoda pri iskanju binarnih sekvenc dolžine 31 in 41. Prva nevronska mreža je učena pravil popačene simetrije. Ta pravila zmanjšajo dimenzijo iskalnega prostora, nevronska mreža pa se jih je uspela naučiti. V želji izboljšanja obstoječega mehanizma smo naučili tudi drugo nevronsko mrežo, ki v iskalnem algoritmu doseže statistično signifikantno boljše rezultate kot mehanizem popačene simetrije. Za to nevronsko mrežo dodatno analiziramo število funkcijskih ovrednotenj za dosego najboljše znane rešitve. Izkazalo se je, da izboljšana nevronska mreža za dosego najboljših znanih rešitev potrebuje manj funkcijskih ovrednotenj kot uporaba pravil popačene simetrije.
Keywords:nevronske mreže, binarne sekvence, iskalni algoritem, samoizogibni sprehod
Place of publishing:Maribor
Place of performance:Maribor
Publisher:[J. Popič]
Year of publishing:2023
Number of pages:1 spletni vir (1 datoteka PDF (IX, 42 f.))
PID:20.500.12556/DKUM-85206 New window
UDC:004.8.032.26(043.2)
COBISS.SI-ID:168580099 New window
Publication date in DKUM:21.09.2023
Views:520
Downloads:111
Metadata:XML DC-XML DC-RDF
Categories:KTFMB - FERI
:
Copy citation
  
Average score:(0 votes)
Your score:Voting is allowed only for logged in users.
Share:Bookmark and 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:22.08.2023

Secondary language

Language:English
Title:Use of neural networks in search of binary sequences with low autocorrelations
Abstract:The usage of neural networks is getting more and more widespread in everyday life as well as in various scientific fields. Regardless of their widespread use, there are still specific research problems for which the usability of neural networks has yet to be tested. One such problem is the search for binary sequences of different lengths with their autocorrelations as small as possible, commonly called a "low-autocorrelation binary sequence problem". Their specific properties can be utilized in many different research topics, but finding them is a complex combinatorial problem. Our novel work presents two neural networks that guide the self-avoiding walk algorithm in the search for sequences of lengths 31 and 41. The first neural network is trained on skew-symmetric sequences, and it was able to mimic this accurately. Additionally, we present another neural network with an improved mechanism that yields statistically significantly better results than skew-symmetry. We also compared the number of function evaluations needed to obtain the best-known results. Self-avoiding walk utilizing an improved neural network used fewer function evaluations than the one utilizing skew-symmetry.
Keywords:neural networks, binary sequences, search algorithm, self-avoiding walk


Comments

Leave comment

You must log in to leave a comment.

Comments (0)
0 - 0 / 0
 
There are no comments!

Back
Logos of partners University of Maribor University of Ljubljana University of Primorska University of Nova Gorica