Campus Event Calendar
Event Entry
What and Who
The Discovery of Slowness
Herr Dr. Laurenz Wiskott
MPI-Kolloquium
AG 1, AG 2, AG 3, AG 4, AG 5
AG Audience
Note: We use this to send email in the morning.
Date, Time and Location
Friday, 28 January 2005
17:30
-- Not specified --
45
HS 3
Saarbrücken
Abstract
In this talk I will present slowness as a relatively new learning principle for the processing of (possibly high-dimensional) time-series; our corresponding algorithm is referred to as 'slow feature analysis' (SFA). The objective is to find input-functions that extract most slowly varying output-signals from the time series (without any low-pass filtering). Slow feature analysis has been developed in context of biological modelling but has also been applied to technical problems. (i) In a hierarchical network as a simple model of the visual system SFA can be used for unsupervised learning of invariances to translation, scaling, etc. for whole objects (here demonstrated for patterns in 1D). (ii) Applied to quasi-natural image sequences (in 2D) SFA learns functions that reproduce many properties of complex cells in primary visual cortex. (iii) SFA can be used to extract slowly varying parameter variations underlying a fast dynamical process, which can be an important step in the analysis of non-stationary time-series. (iv) In a further development we have combined SFA with independent component analysis (ICA) resulting in an algorithm we refer to as independent slow feature analysis (ISFA), which is able to perform nonlinear blind source separation.
Contact
--email hidden
passcode not visible
logged in users only
Bahareh Kadkhodazadeh, 01/25/2005 10:59 -- Created document.