好吧,这有点涉及,但你可以获得所有最大值的索引,然后使用randi和accumarray随机选择一个:
%# (1) Find the maxima
%# if you are interested in the global maximum
%# that may occur multiple times along dimension 6
[maxVal,maxIdx] = max(Defender.Q(:));
%# ALTERNATIVELY
%# if you are interested in local maxima along dimension 6
maxVal = max(Defender.Q,[],6);
maxIdx = find(bsxfun(@eq,Defender.Q,maxVal));
%# (2) pick random maximum for each 5D subarray
%# this assumes that there is no dimension #7 etc
%# In case there is, you need to add a column of ones
%# and then d7 etc to second input of accumarray
%# find row, col, etc subscripts of the maxima
[d1,d2,d3,d4,d5,d6] = ind2sub(size(Defender.Q),maxIdx);
%# create a 5-d array, containing one random index
%# from the maxima along dimension 6, or NaN
randIdx = accumarray([d1,d2,d3,d4,d5],d6,[],@(x)x(randi(length(x))),NaN);