Primjena statističkog učenja na proširenu semi-empirijsku formulu mase

master's thesis

• Title (english): Application of statistical learning to expanded semi-empirical mass formula
• Author: Mihaela Bezak
• Mentor: Nils Paar (mentor)
• Committee member: Nils Paar (predsjednik povjerenstva)
• Committee member: Matko Milin (član povjerenstva)
• Committee member: Davor Horvatić (član povjerenstva)
• Committee member: Tamara Nikšić (član povjerenstva)
• Committee member: Mihael Makek (član povjerenstva)
• Granter: University of Zagreb; Faculty of Science; (Department of Physics); Zagreb
• Defense date and country: 2019-06-06, Croatia
• Scientific / art field, discipline and subdiscipline: NATURAL SCIENCES; Physics

Abstract

U nuklearnoj fizici već više od osamdeset godina u primjeni je Bethe–Weizsäcker formula mase (BW). BW formula sadrži osnovne elemente koji su dovoljni za kvantitativan opis energija vezanja. Unatoč postojanju modernih, mnogo kompleksnijih modela mase, zbog svoje jednostavnosti i intuitivne povezanosti s fizikom atomske jezgre BW formula je i dalje interesantna tema istraživanja. Motiv ovog rada je opisivanje karakteristika BW formule uporabom podataka izmjerenih energija vezanja. Metode statističkog učenja u tome predstavljaju vrlo efikasan alat koji može uvidjeti pravilnosti te značajnost pojedinih doprinosa. Dobiva se zanimljiv rezultat da najznačajniji doprinos energiji vezanja imaju osnovni članovi iz BW formule, dok novo pridodani članovi predstavljaju korekcije manjeg doprinosa na osnovnu formulu. Dodavanje novih članova sustavu opravdano je samo ako novo pridodan član sa sobom nosi novi fizikalni sadržaj, koji prethodno nije bio uključen u modelu. Povećavanjem kompleksnosti, povećavamo varijabilnost na štetu generalizacije što je prikazano pomoću train-test pogreške.

Abstract (english)

Bethe–Weizsäcker (BW) formula has been used by nuclear physicists for more than eighty years. The BW formula contains the basic terms sufficient for a quantitative description of nuclear binding energies. Although more modern and complex mass models exist, the BW formula is still widely used and studied due to its simplicity and intuitive connection with atomic physics. The aim of this thesis is to describe characteristics of the BW formula by using experimentally observed binding energies. Statistical learning methods are shown to be a very effective tool for identifying patterns and the weights of contributing terms. We come to conclusion that the most relevant contribution to binding energies comes from the basic terms of the BW formula, and any additional terms only give small corrections. Adding a new term to the system is only justified if it introduces physics ignored by the basic model. Adding complexity also increases variability at the expense of generalisation, as it is shown by the train-test error.

• Keywords: Bethe–Weizsäcker formula mase; proširena Bethe–Weizsäcker formula mase; primjena statističkog učenja
• Keywords (english): Bethe–Weizsäcker mass formula; statistical learning; improved BW formula
• Language: croatian
• URN:NBN: urn:nbn:hr:217:224872
• Study programme: Title: Physics; specializations in: Research; Course: Research; Study programme type: university; Study level: integrated undergraduate and graduate; Academic / professional title: magistar/magistra fizike (magistar/magistra fizike)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Open access
• Created on: 2019-09-05 11:28:49