20.500.12556/DKUM-95729 Napovedovanje lastnosti molekul z grafovskimi nevronskimi mrežami magistrsko delo Predicting molecular properties with graph neural networks Grafovske nevronske mreže (GNN) so se izkazale kot zmogljivo orodje za napovedovanje lastnosti struktur, predstavljenih v obliki grafov. Molekule lahko predstavimo kot grafe, kjer vozlišča predstavljajo atome, povezave pa kemijske vezi med njimi. V magistrski nalogi preučujemo učinkovitost GNN v kemoinformatiki za napovedovanje topnosti in temperature vrelišča molekul. Napovedovanje izvajamo z večopravilnim modelom ter z ločenimi, specializiranimi modeli za posamezno lastnost. Graph Neural Networks (GNN) have proven to be a powerful tool for predicting the properties of structures represented as graphs. Molecules can be represented as graphs, where nodes represent atoms and edges represent the chemical bonds between them. In this master’s thesis, we investigate the effectiveness of GNN in chemoinformatics for predicting molecular solubility and boiling point. The predictions are performed using a multitask model as well as separate, specialized models for each property. grafovske nevronske mreže graf SMILES napovedovanje lastnosti molekul graph neural networks graph SMILES prediction of molecular properties true true false [L. Novak] Slovenski jezik Angleški jezik Magistrsko delo/naloga 2025-10-16 11:43:36 2025-12-16 08:58:15 2025-12-23 03:18:27 0000-00-00 00:00:00 2025 0 Maribor 0 Maribor 1 spletni vir (1 datoteka PDF (VIII, 48 str.)) 0000-00-00 NiDoloceno NiDoloceno NiDoloceno 0000-00-00 0000-00-00 0000-00-00 004.8.032.26(043.2) 262461955 MAG_Novak_Lovro_2025.pdf MAG_Novak_Lovro_2025.pdf 1 6DC0B0E973EF445BAAEC1A87D5F3CE4B d6122b098d4638021c70b36c336fb79f327efccce3325bcf2c7fdfea414a96aa ccff3e4e-ab36-11f0-a0a8-00155d000105 https://dk.um.si/Dokument.php?lang=slv&id=199961 Fakulteta za elektrotehniko, računalništvo in informatiko 0 0 0