"Machine Learning and Artificial Neural Networks: The Promising Approaches in Data Analysis, Simulations, and Modeling of Materials" başlıklı sunum 5 Ekim 2016 tarihinde saat:14:00'da 4. kat doktora tez savunma salonunda Dr. Berk Onat tarafından verilecektir.
Machine learning and artificial neural networks (ANN) have been the active research fields in data analysis and image recognition. While many of the studies are concentrated on deep learning in image recognition through the last decade, the need to analyse big data in other fields such as signal processing, social networks, finance, biological systems, material and drug design draw an immense attraction to new model development in machine learning. In this talk, I will present a new model for ANNs that can capture the interactions without a bias to showcase the impact of machine learning in materials modelling.
The new ANN architecture can be efficiently use in realistic simulations of Li-ion batteries with Si anodes where first-principle approaches are inaccessible in large systems and long time-scales. After explaining the training procedures and the new memory embedding technique in ANNs, the rest of my talk will focus on how ANN can learn the interactions in a material by means of probing the mind of the network with PCA and t-SNE methods. I will finish my talk by discussing several other potential uses of machine learning and memory embedding method with ANNs in various fields.
Dr. Berk Onat received his MSc and PhD degree in Computational Science and Engineering from Informatics Institute at Istanbul Technical University. He has BSc degree in Physics Engineering from Istanbul Technical University. In his PhD, he developed interatomic potentials using adaptive particle swarm optimisation within the context of parallel computing of the fuzzy logic scheme. After his PhD, he continued working on determining thermodynamical properties of metal alloys at Sabancı University as a postdoctoral researcher.
Following his first postdoc in 2014, he moved to USA for his second postdoc in Professor Efthimios Kaxiras' group at School of Engineering and Applied Sciences of Harvard University and worked on machine learning approaches in computational materials design.
Since June 2016, Dr. Onat is working with Dr. Engin Durgun and Asst. Prof. Çoşkun Kocabağ at UNAM-National Nanotechnology Research Center of Bilkent University. His recent post-doctoral works are based on developing machine learning approaches for the simulations of two-dimensional materials.