dna2vec: vektorska reprezentacija k-torki različite duljine

undergraduate thesis

• Title (english): dna2vec: vector representations of variable-length k-mers
• Author: Mateo Kutnjak
• Mentor: Mile Šikić (mentor)
• Committee member: Mile Šikić (predsjednik povjerenstva)
• Committee member: Krešimir Križanović (član povjerenstva)
• Committee member: Robert Vaser (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2017-07-10, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Usporedba dijelova genoma obavlja se prilagođavanjem
metode word2vec radu s k-torkama. Ispitivana metoda dna2vec omogućuje rad s varijabilnom duljinom k-torki u itervalu [3, 8]. Treniranje izvedbe Skip-Gram koja u obzir uzima kontekst genoma oko k-torke, dobija se pozicija k-torke u višedimenzionalnom vektorskom prostoru. Kosinusna udaljenost dvije vektorske reprezentacije k-torki daje mjeru njihove sličnosti. Ispitana je i dokazana korelacija kosinusne udaljenosti metodom dna2vec i rezultata algoritma Needleman-Wunsch za par k-torki. Dodatno su prodiskutirani razlozi dugotrajnog učenja te su predložena ubrzanja učenja te poboljšanje efikasnosti ispitanog modela.

Abstract (english)

Adjustment of known method word2vec is used for comparing parts of genome. Innovation of method dna2vec is possibility to work with variable length k-mers. Training of Skip-Gram, the word2vec derivation, which takes context of k-mers, results with multidimensional representation of k-mers. Cosine distance of two k-mer vector representations gives similarity measure. In this paper correlation between dna2vec cosine distance and Needleman-Wunsch score was tested and proven. Reasons behind
prolonged training were discussed and possible solutions for faster learning and better efficiency were submitted.

• Keywords: dna2vec; k-torke; sličnost; strojno učenje; nouronska mreža
• Keywords (english): dna2vec; k-mers; similarity; neural network; machine learning
• Language: croatian
• URN:NBN: urn:nbn:hr:168:814254
• 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: 2019-03-13 22:11:10