Optimiranje strategija za igru funkcija uz pomoć simboličke regresije

undergraduate thesis

• Title (english): Optimization of function game strategies with symbolic regression
• Author: Mislav Jelić
• Mentor: Domagoj Jakobović (mentor)
• Committee member: Domagoj Jakobović (predsjednik povjerenstva)
• Committee member: Josip Jukić (član povjerenstva)
• Committee member: NIna Drobac (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2023-07-05, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Ovaj rad je opisao što su evolucijski algoritmi, neke vrste evolucijskih algoritama te kako i gdje se koriste, te demonstrirao njihovu uporabu u rješavanju problema aproksimacije funkcija preko igre funkcija. Strategije za aproksimiranje funkcije bile su optimizirane oslanjajući se na simboličku regresiju, a populacije su bile popunjene jedinkama građenim od različitih vrsta modela. Pokazalo se da je genetsko programiranje bio najuspješniji model te samom svojom građom prikladan za aproksimaciju funkcija, jer su potencijalna rješenja koje GP vraća upravo funkcije. Problem odabira ulaznih parametara bio je riješen sa aktivnim učenjem, gdje su se nakon prvih n parametara svi ostali birali principom aktivnog učenja, što je poprilično ubrzalo aproksimiranje funkcija. Pokazano je da kombinacija evolucijskih algoritama sa aktivnim učenjem može značajno poboljšati aproksimiranje funkcija.

Abstract (english)

This paper described what evolutionary algorithms are and explained some of the types of evolutionary algorithms. It is also mentioned how and where they are used, and their use is demonstrated in solving the problem of approximation of functions through the game of functions. Strategies for approximating the function were optimized relying on symbolic regression, and populations were filled with individuals constructed from different types of models. It turned out that genetic programming was the most successful model and, by its very structure, suitable for the approximation of functions, because the potential solutions returned by GP are precisely functions. The problem of selecting input parameters was solved with active learning, where after the first n parameters, all others were selected by the principle of active learning, which considerably accelerated the approximation of functions. It is shown that the combination of evolutionary algorithms with active learning can significantly improve the approximation of functions.

• Keywords: evolucijski algoritmi; aproksimacija funkcija; simbolička regresija; aktivno učenje; genetsko programiranje
• Keywords (english): evolutionary algorithms; approximation of functions; symbolic regression; active learning; genetic programming
• Language: croatian
• URN:NBN: urn:nbn:hr:168:864629
• Study programme: Title: Computing; Study programme type: university; Study level: undergraduate; Academic / professional title: sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva (sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2023-10-11 09:09:38