Logistic Regression with a Laplacian prior on the Eigenvalues: Convex duality and application to EEG classification
Add to your list(s)
Download to your calendar using vCal
If you have a question about this talk, please contact Zoubin Ghahramani.
Note different room and time
We propose a matrix coefficient logistic regression for the classification of
single-trial ElectroEncephaloGraphy (EEG) signals. The method works in the
feature space of all the variances and covariances between electrodes.
The problem is formulated in a single convex optimization problem with the
spectral $\ell_1$-regularization of the coefficient matrix. In addition, we propose
an efficient optimization algorithm based on a simple interior-point method.
The convex duality plays the key role in this implementation.
Classification results on 162 Brain-Computer Interface (BCI) datasets
show significant improvement in the classification accuracy against $\ell_2$-regularized
logistic regression, rank=2 approximated logistic regression as well as
Common Spatial Pattern (CSP) based classifier, which is a popular technique
in BCI . Connections to LASSO , GP classification with a second order
polynomial kernel, and SVM are discussed.
This talk is part of the Machine Learning @ CUED series.
This talk is included in these lists:
Note that ex-directory lists are not shown.
|