Problem

In fuse_lfp:L137 the $\text{norm}$ of the leadfield that has been reduced from $sensors\times vertices$ to $sensors\times ROIs$ does not correspond to any known $\text{norm}$.

More precisely, the $\text{norm}$ is estimated in code as: sqrt(trace(CL{m}*CL{m}')/Nchn(m)),
but, for matrices $\text{Frobenius norm}$* is defined as sqrt(trace(CL{m}*CL{m}')) (of many sources, here's one at MATLAB's docs)
while $L_2 \text{ norm}$ for a matrix (also called the spectral norm) is defined as max(svd(CL{m})) (MATLAB's docs)

*For matrices: $\text{Frobenius norm}$ = norm(CL{m}, 'fro') = sqrt(trace(CL{m}*CL{m}')) = sqrt(sum(CL{m}(:).^2))

Reproducible example A

x = ones(5);

% Frobenius norms
norm(x, 'fro')      % returns 5
sqrt(trace(x*x'))   % returns 5
sqrt(sum(x(:).^2))  % returns 5

% L2 norms
norm(x, 2)          % returns 5
max(svd(x))         % returns 5

% Implemented norm
sqrt(trace(x*x')/5) % returns 2.2361

Reproducible example B

x = magic(5);

% x =
%
%    17    24     1     8    15
%    23     5     7    14    16
%     4     6    13    20    22
%    10    12    19    21     3
%    11    18    25     2     9

% Frobenius norms
norm(x, 'fro')      % returns 74.330
sqrt(trace(x*x'))   % returns 74.330
sqrt(sum(x(:).^2))  % returns 74.330

% L2 norms
norm(x, 2)          % returns 65
max(svd(x))         % returns 65

% Implemented norm
sqrt(trace(x*x')/5) % returns 33.242

Solution

As the implemented $\text{norm}$ closely matches the $\text{Frobenius norm}$, switch to MATLAB's default implementation with norm(x, 'fro')