Uczenie się dynamiki liniowej na podstawie częściowych obserwacji - porażka, zrozumienie 
i sukces w projektach naukowych
[Learning linear dynamics from partial observations - failure, understanding and success 
in scientific projects]

Mateusz Wiliński (Theoretical Division, Los Alamos National Laboratory, USA & 
Faculty of Information Technology and Communication Sciences, Tampere University, Finland)


Linear models are broadly used in describing processes from different fields. Not only are 
they easy to understand, interpret and apply, but they can also easily be learnt from data 
by regression. But what if the data is limited and only part of the system is observed? Can 
interdependence between observed and hidden part allow us to recover the full model and not 
only its visible subpart? Under what conditions this can be done and how? Apart from 
answering the above questions I will also share the path, which lead us to the solution. 
A rough path, so common in many areas of science -- from failure, through understanding, 
to success.