I have an Engineering degree - Major in Data Science (obtained from École des Pont ParisTech, France) and a MSc in Applied Mathematics and Informatics (obtained from NSU, Russia).
Currently, I’m enrolled in PhD at Inria Méditerranée supported by the French government, through the 3IA Côte d’Azur Investments. I’m pleased to present a short description of my PhD and other projects where I’ve participated during my studies.
I started my PhD, named Learning Cardiac 3D Electromechanical Dynamics with PDE-based Physiological Constraints for Data-Driven Personalized Predictions in Cardiology, in march 2020 under supervision of Maxime Sermesant (Inria, Epione) and Patrick Gallinari (Sorbonne University, LIP6, MLIA).
The scientific objective of this project is to combine the advantages of biophysics and deep learning methods, and to develop hybrid models exploiting the complementarity of the two approaches. The main idea is to introduce physiological priors in learning systems through biophysical modelling by learning spatiotemporal dynamics from simulations and by introducing physically motivated constraints relative to these dynamics.
The objective is to exploit optimally the large amounts of data available in this field together with well-known properties of biophysical cardiac dynamics. Besides, this would also enable to propose a datadriven correction of biophysical models error.
Passioned by computer vision and the applications of GANs and VAEs for image translation and synthesis I’ve also participated in the different (non deep learning) projects to get a deep understanding of basic image processing. More than that, I have a strong background in math and I’ve been doing research on global optimization methods for multi-dimensional functions for several years.
The list of my projects is below :
True face detection (for Ubble.ai)
CLIVIO Oscar, CORVISIER Jean-Christophe, KASHTANOVA Victoriya and LOISEAU Romain
/ code / poster
Solution development for real user authentification against various types of digital fraud, like : high-resolution photos or videos, 3D masks etc.
CycleGAN analysis (Research project in DL)
KASHTANOVA Victoriya, PESQUEREL Fabien
Deep investigation of the CycleGAN model paper : construction, training configurations, used metrics etc. Also studying the impact of the “anchor loss” (we added one matched pair of images and its loss to the unpaired datasets) on the model performance and the ability of the CycleGAN to do a transfer learning.
Deep Reinforcement Learning agents playing DOOM (Research project in RL and Computer Vision)
HURAULT Samuel and KASHTANOVA Victoriya,
Investigation and optimization of two deep Reinforcement Learning frameworks which learn sensorimotor control in partially observable 3D environments.
Computer vision for serial numbers (for Pzartech.com)
BOZOU Sacha, DE SOUSA Philippe, DUMONT Louis and KASHTANOVA Victoriya
/ demo / code
Solution development for serial number recognition on the mechanical parts.
Automatic machine translation (Research project in mathematics)
FERREIRA DE SOUSA Philippe, KASHTANOVA Victoriya and SOUSSI Nada, supervised by MARLET Renaud (IMAGINE Lab)
/ demo / code / poster
Population based epidemiological modelling (Research project in mathematics)
KASHTANOVA Victoriya, supervised by KABANIKHIN Sergey and KRIVOROTKO Olga
/ arxiv / code
Knowing that the models of the epidemic spread could be described by systems of nonlinear ODEs whose coefficients characterize the features of population and disease.I worked on the construction and investigation of numerical algorithms for determining these coefficients by the global minimization of the multidimensional target function.