Solvers¶

invertmeeg provides 84 inverse solvers organized into 9 categories.

Each solver has a full name and a stable solver id. Instantiate solvers with Solver(solver_id):

from invert import Solver

solver = Solver("MNE")

Categories¶

Minimum Norm ¶

15 solvers

Minimum norm methods estimate source activity by finding the solution with the smallest norm (typically L2) that explains the measured data.

Bayesian ¶

14 solvers

Bayesian methods use probabilistic frameworks to estimate source activity, incorporating prior knowledge about source distributions.

Beamformers are spatial filters that estimate source activity at each location independently by constructing filters that pass signals from the target location while suppressing contributions from other sources.

Subspace Methods ¶

8 solvers

Subspace methods (including MUSIC variants) exploit the orthogonality between signal and noise subspaces to localize sources.

Matching Pursuit ¶

10 solvers

Matching pursuit and related greedy algorithms iteratively select sources that best explain the residual signal.

Dipole Fitting ¶

2 solvers

Dipole fitting methods model brain activity as a small number of equivalent current dipoles and estimate their locations and orientations by minimizing the difference between measured and predicted signals.

Neural Networks ¶

5 solvers

Neural network-based methods use deep learning to learn the inverse mapping from sensor data to source activity.

Hybrid ¶

2 solvers

Hybrid methods combine multiple inverse approaches to leverage their respective strengths, such as using beamformers for initial localization followed by sparse reconstruction.

Baseline ¶

1 solver

Baseline methods provide reference implementations for benchmarking purposes, such as random noise generators for null hypothesis testing.