Academics
How-to: Add a News Item
Seminar: Microsoft Live, Software, and Studios (9/12/2008)
Seminar: Yaroslav Halchenko, "PyMVPA: Fathom Brain Function through Multivariate Pattern Analysis" (4/17/2009)
Friday, 4/17, 1:15 - 2:15, POST 127
Seminar Series in Machine Learning and Computational Biology:
PyMVPA: Fathom Brain Function through Multivariate Pattern Analysis
Yaroslav Halchenko
Rutgers University
Friday, 4/17, 1:15 - 2:15,
POST 127
PyMVPA: Fathom Brain Function through Multivariate Pattern Analysis
Yaroslav Halchenko
Rutgers University
Friday, 4/17, 1:15 - 2:15,
POST 127
In the last five decades the number of techniques available for non-invasive functional brain imaging has increased dramatically. Researchers today can choose from a variety of neural data modalities such as fMRI, EEG, MEG, etc. Only recently neuroimaging researchers have begun to explore various multivariate methods to address the shortcomings of the conventional analysis approaches.
By relying on existing statistical learning methods, we offer a methodology for the reliable analysis of neural data at different levels of investigation. The proposed methodology has been formalized into PyMVPA: a free and open source Python framework for statistical learning based data analysis. Drawing on the field of statistical learning theory, analysis techniques available in PyMVPA posses explanatory power that could provide new insights into the functional properties of the brain. By constructing a decoder of neural data new analysis methods often reverse the direction of the analysis and target the description of the behavior or the environment in terms of the registered neural activity. Such reversed paradigm can account for the covariance/causality structure within the data and often allows for single-trial analysis of various neural data modalities by decoding them into variables of interest. Unbiased testing of the decoder on new samples of the data provides an assessment of decoding performance validity. Furthermore, consecutive analysis of the constructed decoder's sensitivity allows to identify neural signal components relevant to the task of interest, hence providing desired localization. These features make decoding approach more powerful than conventional hypothesis-testing univariate methods which currently dominate the neuroscience field.
Please visit http://www.pymvpa.org to learn more about the project.
