Official URL: http://www.bcl.hamilton.ie/~barak/papers/tr-unm-cs...

Iterative gradient methods like Levenberg-Marquardt (LM) are in widespread use for source localization from electroencephalographic (EEG) and magnetoencephalographic (MEG) signals. Unfortunately LM depends sensitively on the initial guess, particularly (and counterintuitively) at higher signal-to-noise ratios, necessitating repeated runs. This, combined with LM's high per-step cost, makes its computational burden quite high. To reduce this burden, we trained a multilayer perceptron (MLP) as a real-time localizer. We used an analytical model of quasistatic electromagnetic propagation through the head to map randomly chosen dipoles to sensor activities, and trained an MLP to invert this mapping in the presence of various sorts of noise. With realistic noise, our MLP is about five hundred times faster than n-start-LM with n = 4 to match accuracies, while our hybrid MLP-start-LM is about four times more accurate and thirteen times faster than 4-start-LM.