University of Birmingham > Talks@bham > Artificial Intelligence and Natural Computation seminars > Probabilistic Principal Component Analysis for Time Series

Probabilistic Principal Component Analysis for Time Series

Add to your list(s) Download to your calendar using vCal

If you have a question about this talk, please contact Hector Basevi.

Host: Prof. Peter Tino

Speaker’s website: https://sites.google.com/site/ngiann/

Abstract: In this talk we present an approach that modifies probabilistic principal component analysis (PPCA) so that it can be applied on a dataset of multiple time series for the purpose of visualisation. An important building block in this approach is the use of Echo State Networks (ESNs). ESNs are recurrent neural networks that possess a randomly “wired” layer and are trained by adapting only the output weights. For each time series in the dataset, we fit an ESN and record the generated hidden states. We explain how with a simple modification we can reformulate the PPCA objective function so that it incorporates the previously recorded hidden states. Incorporating the hidden states is significant as it allows us to take into account the temporal behaviour of the time series. We demonstrate the significance of this on a few benchmark problems and on a real dataset.

This talk is part of the Artificial Intelligence and Natural Computation seminars series.

Tell a friend about this talk:

This talk is included in these lists:

Note that ex-directory lists are not shown.

 

Talks@bham, University of Birmingham. Contact Us | Help and Documentation | Privacy and Publicity.
talks@bham is based on talks.cam from the University of Cambridge.