University of Birmingham > Talks@bham > Artificial Intelligence and Natural Computation seminars > Mesh Learning: A functional Brain Network for Modeling the Cognitive Processes*

Mesh Learning: A functional Brain Network for Modeling the Cognitive Processes*

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

If you have a question about this talk, please contact Leandro Minku.

Host: Prof. Ales Leonardis

One of the most challenging problem in studying the human mind is the representation of information coded in the brain. How is information represented and how does the representation change depending on the type and nature of the information? This has long been an intriguing and challenging question in the history of science across many disciplines.

In this study, we construct a functional brain network and use this network for modeling and recognizing the cognitive processes. The nodes of the suggested network are the voxel intensity values of the functional Magnetic Resonance Images (fMRI). The arc weights are estimated by using the functional neighborhood concept. In order to define functional neighborhood, the similarities between the time series recorded for voxels are measured and functional connectivity matrices are constructed. Then, a local mesh around each voxel is formed by including the p-functionally closest neighboring voxels in the mesh. The arc weights for each mesh are estimated by using a linear regression model. The arc weights are, then, used to train a learning machine.

The proposed method, called Mesh Learning, is tested on a recognition memory experiment, using the fMRI data pertaining to encoding and retrieval of words belonging to ten different semantic categories. A popular classifier, namely k-Nearest Neighbor is trained in order to recognize the semantic category of the item being retrieved based on activation patterns during encoding. The classification performance of the Mesh Learning model, which range in 65-85% is superior to the classical multi-voxel pattern analysis (MVPA) methods, which range in 25-48%, for ten semantic categories.

- This work is supported by Google Inc. and National Research Council of Turkey.

- Research team includes: Mete Ozay, Itir Onal, Orhan Firat, Emre Aksan, Burak Velioglu, Sarper Alkan, ─░lke Oztekin and Uygar Oztekin.

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.