Classification of 1D-Signal Types Using Semi-Supervised Deep Learning

U okviru ovog rada, predložili smo metode za detekciju tipova grafova pokrivenosti na temelju polu-nadziranih modela dubokog učenja: autoenkodera i generativnih suparničkih mreža. Testirali smo performanse svih modela koristeći bazu koja sadržava očitanja s više referentnih genoma. Ručno smo označili dio podataka i usporedili rezultate s obzirom na količinu označenih primjera koji su bili korišteni prilikom treniranja. Ovu detekciju smo ugradili u proces sastavljanja i dobili dobre rezultate. Izvorni kod je dostupan na https://github.com/tomislavsebrek/diplomski.In this thesis, we proposed methods for detecting the type of the coverage graph based on semi-supervised deep learning models: autoencoders and generative adversarial networks. We evaluated the performance of each model based on the dataset that contained reads from multiple reference genomes. We have manually labeled some of the data and compared the results of all models with respect to the number of labeled examples that were used during the training process. We have embedded this detection in the assembly process and achieved good results. The source code is available at https://github.com/tomislavsebrek/diplomski