Minimum-Norm Source Localization

Invert a noisy EEG lead matrix with Tikhonov regularization, inspect resolution and L-curves, and expose source-model mismatch.

Chapter 7
MNE
Tikhonov
inverse problem

Residual norm

1.807 µV

Source norm

0.475

Peak index / truth

39 / 20

Unscaled source RMSE

0.160

Source estimate without amplitude rescaling

True unit sourceMNE estimateResolution point-spread

L-curve (no oracle corner selection)

Forward fit can conceal source errors

MeasuredFitted

Model scope

Underdetermined infinite-medium EEG lead matrix with south-pole reference; source unit = 10 nA·m, data = µV. Equal independent channel-noise model; no covariance whitening or depth weighting. λ uses this fixed scaled basis and λ² penalty. Ground-truth radius mismatch tests model error; the L-curve is a trade-off, not an automatically optimal or clinically validated corner.

Preparing the simulation…

Conceptual bridges & related modules

Lead Fields, Reciprocity & Sensitivity
prerequisite · Chapter 7
Invert the same scaled source basis and inspect its unresolved combinations.
MEG vs EEG: The Sarvas Dipole Model
companion · Chapter 7
Distinguish forward source visibility from inverse localization validity.
Bioelectric Uncertainty, Stimulation & EIT
companion · Chapter 7
Compare EEG and discrete EIT regularized inversions and model mismatch.
Spectral Segmentation Explorer
companion · Chapter 9 · Section 9
Compare ill-conditioned source estimation with image-domain delineation.