Μαργαρίτης, ΚωνσταντίνοςΧριστοφορίδης, ΑριστείδηςΔιπλωματική εργασία--Πανεπιστήμιο Μακεδονίας, Θεσσαλονίκη, 2021.Approved for entry into archive by Κυριακή Μπαλτά (balta@uom.gr) on 2021-07-13T17:11:27Z (GMT) No. of bitstreams: 3 license_rdf: 701 bytes, checksum: 42fd4ad1e89814f5e4a476b409eb708c (MD5) ChristoforidisAristeidisMsc2021.pdf: 1601827 bytes, checksum: 8d26bffdbf7a2c4b61c010d30f6569b8 (MD5) ChristoforidisAristeidisMsc2021present.pdf: 1616687 bytes, checksum: 898fcf57f6202e038d656de1100ee3bf (MD5)Made available in DSpace on 2021-07-13T17:11:27Z (GMT). No. of bitstreams: 3 license_rdf: 701 bytes, checksum: 42fd4ad1e89814f5e4a476b409eb708c (MD5) ChristoforidisAristeidisMsc2021.pdf: 1601827 bytes, checksum: 8d26bffdbf7a2c4b61c010d30f6569b8 (MD5) ChristoforidisAristeidisMsc2021present.pdf: 1616687 bytes, checksum: 898fcf57f6202e038d656de1100ee3bf (MD5) Previous issue date: 2021-06-29In this thesis, we propose a new neural architecture search algorithm that performs network discovery in global search spaces. We introduce a novel network representation that organizes the topology on multiple hierarchical levels of varying abstraction and develop an evolution based search process that exploits this structure to explore the search space. Our approach involved a curation system that selects well performing network components and uses them in subsequent generations to build better networks. Next, we investigate how the proposed method performs on different types of data. First, we apply our method on an activity recognition time series dataset and manage to discover a topology with impressive performance. We also test the method on two image classification datasets, Fashion-MNIST and NAS-Bench-101 and achieve accuracies of 93.2% and 94.8% respectively in a small amount of time.Submitted by ΑΡΙΣΤΕΙΔΗΣ ΧΡΙΣΤΟΦΟΡΙΔΗΣ (aid20003@uom.edu.gr) on 2021-07-13T08:25:04Z No. of bitstreams: 3 license_rdf: 701 bytes, checksum: 42fd4ad1e89814f5e4a476b409eb708c (MD5) ChristoforidisAristeidisMsc2021.pdf: 1601827 bytes, checksum: 8d26bffdbf7a2c4b61c010d30f6569b8 (MD5) ChristoforidisAristeidisMsc2021present.pdf: 1616687 bytes, checksum: 898fcf57f6202e038d656de1100ee3bf (MD5)http://dspace.lib.uom.gr/handle/2159/25659Πανεπιστήμιο ΜακεδονίαςCC0 1.0 Παγκόσμιαhttp://creativecommons.org/publicdomain/zero/1.0/Deep neural networksNeural architecture searchA novel evolutionary algorithm for hierarchical neural architecture searchElectronic Thesis or DissertationText2021-07-13T17:11:27ZΤεχνητή νοημοσύνηArtificial Intelligenceisi - Computer Science, Artificial Intelligence covers resources that focus on research and techniques to create machines that attempt to efficiently reason, problem-solve, use knowledge representation, and perform analysis of contradictory or ambiguous information. This category includes resources on artificial intelligence technologies such as expert systems, fuzzy systems, natural language processing, speech recognition, pattern recognition, computer vision, decision-support systems, knowledge bases, and neural networks.Μεταπτυχιακή εργασίαMaster thesisΠανεπιστήμιο ΜακεδονίαςGreek Aggregator OpenArchives.gr | National Documentation Centre (EKT)