Keynote Speakers
Neuroinformatics - 2017
Видеозаписи лекций и пленарных докладов
Monday, October 2 11:20 – 13:00
Lecture-hall Алексеевский зал
Chair: Prof. DUNIN-BARKOWSKI WITALI
University of Leicester, Great Britain
Non-iterative, reversible and non-destructive learning in multidimensional world
2. DUNIN-BARKOVSKY V.L
The Moscow Institute of Physics and Technology (State University)
Toward and Beyond Human-Level AI
Monday, October 2 14:00 – 16:30
Lecture-hall Алексеевский зал
Chair: Prof. DUNIN-BARKOWSKI WITALI
State Research Center of Russian Federation "Troitsk institute for innovation & fusion research", Moscow
Scripts, symmetry, network
Narrative and scripts are examples of big data studied now with deep learning networks. These types of data contain information reflecting brain functioning and can be used in studies of memory and creativity. Using theater metaphor we consider such topics as symmetrical and asymmetrical logics --- bi-logic and also the approaches connecting bi-logic with ultrametric data analysis. Also we consider the connection of this two topics from the perspective of neural networks theory.
4. STANKEVICH L.A.
Peter the Great St. Petersburg Polytechnic University
Cognitive behavioral system
In the given study development ways of cognitive behavioral systems are discussed. The special attention is taken away to consideration cognitive processes of formation of the behaviors, existing architectures of behavioral systems and construction possibility on their basis cognitive behavioral systems. One of directions of development of such systems is connected with working out suitable for intellectual robot cognitive means of formation of behavior. It is shown that working out of such means is actual, as they can essentially raise efficiency of modern humanoid robots, which have not only anthropomorphic form, but should show humanlike behavior. Problems of developing algorithms and programs for formation of difficult behavioral acts are discussed. Variants of formation of individual and collective behaviors of robots are considered. It is shown that comprehensible behavior for humanoid robots can be attained with use neurological and neuromorphic means. Perspective directions of researches on perfection of navigating behavior by application of cognitive mapping, and also on cognitive development with reference to humanoid robots are discussed.
5. LEMPITSKIY V.S.
Skolkovo Institute of Science and Technology
Towards realistic neural image synthesis
The talk will discuss an emerging area of neural image synthesis, i.e. a class of methods concerned with deep neural networks that can generate realistic images either unconditionally or being conditioned on some information, such as a different image. Neural image synthesis has the potential to revolutionize computer graphics and image editing, while also representing an interesting and important frontier for research in artificial intelligence. The talk will briefly review the most popular computational models for such synthesis. It will then cover some recent approaches for neural image synthesis developed within Skoltech Computer Vision group.
