Thesis supervisor: Gábor Farkas
Location of studies (in Hungarian): ELTE IK, Szombathely Abbreviation of location of studies: IKSEK
Description of the research topic:
Over the last few decades, all fields of life (industry, agriculture, health, etc.) have accumulated data sets that contain a lot of valuable hidden information. Methods to extract these have been developed in many areas of artificial intelligence, such as data mining, machine learning, neural networks. In experimental conditions, on sample databases, these methods perform well, but not so well on data from real environments. There are also cases where the industrial partner poses a completely new question that has no method to answer.
Specifically, in this topic we investigate how to use graph or probability theoretical results to design neural networks that answer predictive questions, or how to use neural networks to compute the error function in matrix factorization processes.
Required language skills: English Recommended language skills (in Hungarian): Hungarian Further requirements: Required pre-studies: Discrete Mathematics I-II, Analysis I-II-III, Linear Algebra, Fundamentals of Data Science, Machine Learning, Parallel and Distributed Algorithms.
High level programming skills, proficiency in HPC (High Performance Computing).