I am thus pleased to announce that a UdeM-McGill-MITACS machine learning seminar will be held this Wesdnesday, Aug 24th. The talk will be given by Benjamin Schrauwen and take place from 15h00-16h00 in the room AA3195 (pavillon Andre-Aisenstadt, click
here for directions) at the Université de Montréal. Abstract and title below. Hope to see you there !
Title: An overview and recent insights on Reservoir Computing
Abstract:
Many practical and industrial applications are temporal in nature: control, model identification, robotics, language processing, speech recognition, biomedical signal processing, ... Reservoir Computing (RC) is a recently proposed Machine Learning technique using untrained recurrent neural networks as feature expansion which allows to train spatio-temporal functionals using only linear regression.
We will introduce the technique and show various tasks on which it has been successfully applied. Recent work will be presented that shows that kernel functions can be defined that effectively simulate infinite sized reservoir systems, and how RC can be seen as fading memory regularizers.
If time permits, we will briefly demonstrate that the ideas of Reservoir Computing can lead to a radically new way of thinking about how computation can be implemented in physical systems.