Interpretabilnost izraza i analiza značajki u simboličkoj regresiji

master's thesis

• Title (english): Expression interpretability and feature analysis in symbolic regression
• Author: Ivona Škorjanc
• Mentor: Domagoj Jakobović (mentor)
• Committee member: Domagoj Jakobović (predsjednik povjerenstva)
• Committee member: Marko Čupić (član povjerenstva)
• Committee member: Marko Đurasević (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2020-02-25, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Predstavljen je problem interpretabilnosti u strojnom učenju s naglaskom na interpretabilnost izraza u simboličkoj regresiji. Objašnjena je potreba za pojednostavljivanjem izraza u simboličkom računanju. Prikazan je pregled dostupnih programskih rješenja za simboličko računanje i način na koji je u njima implementirano pojednostavljivanje. Prikazana je implementacija sustava za pojednostavljivanje prilagođenog okruženju ECF uz primjere. Rezultati su prikazani na nekoliko različitih problema te je zaključeno da je sustav koristan kod analize i ubrzavanja kompleksnijih problema poput problema raspoređivanja.

Abstract (english)

The paper describes the problem of interpretability in different machine learning models with an emphasis on symbolic regression. The need to simplify expressions in symbolic computation is illustrated. A number of relevant existing solutions is listed along with an analysis of their simplification methods. A custom simplifier was developed for easy integration into ECF framework and its usage is demonstrated with examples. Results for several different problems are shown with a conclusion that a simplifier like this is useful for an analysis and acceleration of complex types of problems like scheduling hyperheuristics.

• Keywords: Genetsko programiranje; strojno učenje; pojednostavljivanje izraza; simbolički izrazi; ECF; C++
• Keywords (english): Genetic programming; machine learning; expression simplifier; symbolic expressions; ECF; C++
• Language: croatian
• URN:NBN: urn:nbn:hr:168:355794
• Study programme: Title: Computing; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra inženjer/inženjerka računarstva (magistar/magistra inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2023-01-13 10:47:54