Thesis Details
Analýza genetické příbuznosti aproximativních obvodů
The goal of this thesis is analyzing a large library of approximate circuits (EvoApproxLib) which was created using an evolutionary algorithm and used as a source of genetic data for the purposes of this thesis. More specifically it is a relatedness search in a file containing 24 912 8-bit approximate multipliers which were created by evolution from six different fully functioning parent implementations of multiplication. Gate counts and existence of 16 specific subcircuits were used as relatedness indicators. Various classifiers for assigning multipliers to one of six classes corresponding to parent implementations were trained based on these indicators. A classification success rate of up to 77% was achieved using said indicators. The results of this work show that combinations of specific subcircuits are a strong indicator for identifying which parent circuit the given approximate circuit comes from.
Cartesian genetic programming, evolutionary algorithms, classification, approximation, approximate computing
Hradiš Michal, Ing., Ph.D. (DCGM FIT BUT), člen
Jaroš Jiří, doc. Ing., Ph.D. (DCSY FIT BUT), člen
Křivka Zbyněk, Ing., Ph.D. (DIFS FIT BUT), člen
Lengál Ondřej, Ing., Ph.D. (DITS FIT BUT), člen
@bachelorsthesis{FITBT24400, author = "Vojt\v{e}ch Krej\v{c}\'{i}k", type = "Bachelor's thesis", title = "Anal\'{y}za genetick\'{e} p\v{r}\'{i}buznosti aproximativn\'{i}ch obvod\r{u}", school = "Brno University of Technology, Faculty of Information Technology", year = 2022, location = "Brno, CZ", language = "czech", url = "https://www.fit.vut.cz/study/thesis/24400/" }