Skip to content
New issue

Have a question about this project? # for a free GitHub account to open an issue and contact its maintainers and the community.

#

By clicking “#”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? # to your account

Correct normalization of STE values #3

Open
wants to merge 2 commits into
base: master
Choose a base branch
from
Open

Correct normalization of STE values #3

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
4d399d5
Correct normalization of STE values
seb-berger Nov 15, 2018
5f4d35d
Add a test function to STE
seb-berger Nov 15, 2018
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
51 changes: 35 additions & 16 deletions EEGapp/custom_stuff/f_nste_new.m
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -37,7 +37,20 @@
[P1, P2, P3, P4] = f_Integer2prob(Int_future(:,1), Int_past(:,1), Int_past(:,2));
sft_T = sum( P1 .* (log2( P1.*P4 ) - log2(P2.*P3)) );
%H_XY = sum ( P1.*(log2(P2) - log2(P1.*P2))); *** SEE BOTTOM OF CODE
H_XY = -sum ( P3./P4.*(log2(P3) - log2(P4)))
%H_XY = -sum ( P3./P4.*(log2(P3) - log2(P4)))

% Bugfix by Sebastian Berger:
% NOTE THAT P1, P2, P3, and P4 ARE *NOT* PROBABILITY DISTRIBUTIONS.
% Due to the implementation trick pulled in f_Integer2prob, P1 here
% needs to be used to weight the sum of logarithms. The code will then
% correctly calculate H(Y_F | Y_P), that is, the conditional entropy
% of the future of the sink channel (Y_F), given its own past(Y_P).

H_XY = -sum ( P1.*(log2(P3) - log2(P4)));

% For testing:
% fprintf('Delta: %g\n', H_XY - cond_ent(dim, Int_future(:, 1), Int_past(:, 1)));

T_s(1) = (T(1) - sft_T)/H_XY;

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
Expand All @@ -58,28 +71,34 @@
[P1, P2, P3, P4] = f_Integer2prob(Int_future(:,2), Int_past(:,2), Int_past(:,1));
sft_T = sum( P1 .* (log2( P1.*P4 ) - log2(P2.*P3)) );
%H_XY = sum ( P1.*(log2(P2) - log2(P1.*P2))); *** SEE BOTTOM OF CODE
H_XY = -sum ( P3./P4.*(log2(P3) - log2(P4)))
%H_XY = -sum ( P3./P4.*(log2(P3) - log2(P4)))
H_XY = -sum ( P1.*(log2(P3) - log2(P4)));
T_s(2) = (T(2) - sft_T)/H_XY;

%*** NSTE didn't give the right output with this line of code
% we redid the derivation and we found out that this line should be
% replaced by the one below (as suggested by Dr. Chang)
display('new version running!')

function ent = cond_ent(dim, future, past)
% Easier to trace/debug, but (by far) less efficient.

% Marginal entropy H(X_P)
symbols = unique(past);
H_x = 0;
for n = 1:numel(symbols)
prob = mean(past == symbols(n));
H_x = H_x - prob * log2(prob);
end

% Joint entropy H(X_F, X_P)
joint = future * 10^dim + past;
symbols = unique(joint);
H_xx = 0;
for n = 1:numel(symbols)
prob = mean(joint == symbols(n));
H_xx = H_xx - prob * log2(prob);
end















% Conditional entropy
ent = H_xx - H_x;