wolffd@0: function s = logsumexpv(a)
wolffd@0: % Returns log(sum(exp(a)))  while avoiding numerical underflow.
wolffd@0: %
wolffd@0: % e.g., log(e^a1 + e^a2) = a1 + log(1 + e^(a2-a1)) if a1>a2
wolffd@0: % If a1 ~ a2, and a1>a2, then e^(a2-a1) is exp(small negative number),
wolffd@0: % which can be computed without underflow.
wolffd@0: 
wolffd@0: % Same as logsumexp, except we assume a is a vector.
wolffd@0: % This avoids a call to repmat, which takes 50% of the time!
wolffd@0: 
wolffd@0: a = a(:)'; % make row vector
wolffd@0: m = max(a);
wolffd@0: b = a - m*ones(1,length(a));
wolffd@0: s = m + log(sum(exp(b)));
wolffd@0: