Abstract:
|
In this talk, we will consider the problem of classification using the Morlet transform for fMRI data. Its use has been widespread in engineering, and it is represents a class of wavelet techniques different from more popular versions (e.g., Haar wavelet, Mexican hat wavelet). In this talk, we describe how it can be cast in the kernel machine framework that is popular in the statistical and data mining communities. This lends itself to implementation using standard statistical tools as well as a unified approach to inference. In addition, we will describe an extension that handles two levels of scaling. The approach will be applied to data from a risk-taking study as well as simulated data. This is joint work with Manish Dalwani.
|