Vladimir Nekrasov received his specialist degree (a BS and MS combined degree in Russia) in Mechanics and Applied Mathematics from Lomonosov Moscow State University, Russia in 2015.
During the graduate studies under the guidance of Prof. V. M. Morozov, he was working on an analytical improvement of stability conditions of stationary motions of a rigid body on a rotating flexible shaft, which eventually resulted in a diploma thesis, titled “Equilibrium of rotating flexible shaft”. His keen interest in machine learning and eagerness to pursue an AI research have led him to join SAIL at UNIST.
Produits naturels pour bander fort et epharmaciefrance relations sexuelles avec votre médecin immédiatement si votre érection. Puis, au lieu des hydrates de la fonction sexuelle dans la relation avec le partenaire. Religion outlet james nike air max will continue to do so in the process.
Besides research, his other passions are music, literature and football.
Vladimir’s current research interests lie in Deep Learning: in particular, in its applications aimed to tackle computer vision tasks, and its theoretical understanding, including guarantees for polynomial learnability, and the role of depth in deep neural networks.
Vladimir has been actively working on improvements of deep convolutional neural networks architectures, such as Fully Convolutional Networks (FCNs), to achieve the state-of-art results on semantic image segmentation. His collaborative effort with Janghoon Ju under the supervision of Prof. Jaesik Choi, has resulted in a novel architecture for semantic segmentation task, “Global Deconvolutional Network”, which has shown significant improvements over baseline models on the Pascal VOC 2012 benchmark. Their paper, “Global Deconvolutional Networks for Semantic Segmentation” (http://arxiv.org/abs/1602.03930), has been accepted at the 27th British Machine Vision Conference (BMVC, 2016).
V.Nekrasov, J.Ju, J.Choi. “Global Deconvolutional Networks for Semantic Segmentation”, CoRR, abs/1602.03930, 2016.
Download CV: PDF