V. Drungilas “Collaborative distributed machine learning on blockchain” doctoral dissertation defense

Author, Institution: Vaidotas Drungilas, Kaunas University of Technology

Science area, field of science: Technological Sciences, Informatics Engineering, T007

Scientific Advisor: Prof. Dr. Evaldas Vaičiukynas (Kaunas University of Technology, Technological Sciences, Informatics Engineering, T007)

Dissertation Defense Board of Informatics Engineering Science Field:
Prof. Dr. Robertas Damaševičius (Kaunas University of Technology, Technological Sciences, Informatics Engineering, T007) – chairperson
Assoc. Prof. Dr. Razvan Bocu (Transilvania University of Brasov, Romania, Technological Sciences, Informatics Engineering, T007)
Prof. Dr. Nikolaj Goranin (Vilnius Gediminas Technical University, Technological Sciences, Informatics Engineering, T007)
Prof. Dr. Tomas Krilavičius (Vytautas Magnus University, Natural Sciences, Informatics, N009)
Prof. Dr. Agnė Paulauskaitė-Tarasevičienė (Kaunas University of Technology, Natural Sciences, Informatics, N009)

Dissertation defense meeting will be at Rectorate Hall of Kaunas University of Technology (K. Donelaičio 73–402, Kaunas)

The doctoral dissertation is available at the library of Kaunas University of Technology (Gedimino g. 50, Kaunas)

Annotation: Collaborative distributed machine learning approaches are constrained by insufficient trust, limitations imposed by sensitive data, and a complex adaptation process for the currently existing machine learning solutions. A method for collaborative privacy-preserving distributed machine learning has been proposed, thus enabling the blockchain network participants to collaborate via the model deployment process. The proposed method quantifies the contributions of the participants and enables the usage of knowledge accumulated on the blockchain network via the weighted ensemble or the knowledge distillation approaches. Proof-of-concept implementation has been developed by using the Hyperledger Fabric private blockchain network architecture, thus demonstrating the feasibility of collaborative distributed machine learning. The proposed method introduces novel blockchain oracle components that enables the reusability of existing machine learning solutions. The completed experiments measured the performance of the proposed blockchain network architecture, tested the proposed collaboration contribution evaluation strategy and measured the performance impact of knowledge distillation approach. For the tested dataset and classifier configurations the proposed method increased the classifier performance by 4.8% and 1.9% when compared to single model approach.