MATLAB新增for循环后计算结果异常问题排查

在MATLAB中新增for level循环后，对应变量的计算结果出现异常。固定level=1时代码运行正常，输出如下：

0
0
0
0
0
38

但新增for level = 1:2循环后，p_M_level第二列结果异常，手动修改level值时结果却正常。

原代码（固定level=1） #

level = 1;
for distance = 1:6
  for jj = 1:kol_real
            disp(jj)
            f_pr1 = 7;
            f_pr2 = 13 + distance;
            
            U1 = U(:, f_pr1)+U(:, f_pr2);%sum of two signals u(t) = sin(2*pi*f_0*t) with different f_0
            add_noise_norm = normrnd(0, 0.2 + (0.1 *level), [N + kol_shift, 1]);
            U_pr_norm = U1 + add_noise_norm;%add some noise
                        
            parfor dd = 1:25
                M_norm(dd, jj) = corr(U_pr_norm, U(:, dd));  % calculate correlation  
            end

            [~, zz_M_norm] = findpeaks(M_norm(:, jj), 'NPeaks', 8,'SortStr','descend'); % find peaks of correlation function
            min_M_norm(1:length(zz_M_norm),jj) =  sortrows(zz_M_norm,'ascend');% write value into two array
%            
  end

 min_M_norm_temp = min_M_norm(1:2,:);
 min_M_norm_temp(:, any(min_M_norm_temp == 0)) = [];%delete if i have one zeros element
 count_M_norm = abs(min_M_norm_temp(1, :) - min_M_norm_temp(2, :)) >= 2;% compare using some condition
 p_M_norm(distance) = sum(count_M_norm);
end

新增level循环后的代码 #

for level = 1:2;
for distance = 1:6
for jj = 1:kol_real
            disp(jj)
            f_pr1 = 7;
            f_pr2 = 13 + distance;
            
            U1 = U(:, f_pr1)+U(:, f_pr2);%sum of two signals u(t) = sin(2*pi*f_0*t) with different f_0
            add_noise_norm = normrnd(0, 0.2 + (0.1 *level), [N + kol_shift, 1]);
            U_pr_norm = U1 + add_noise_norm;%add some noise
                        
            parfor dd = 1:25
                M_norm(dd, jj) = corr(U_pr_norm, U(:, dd));  % calculate correlation  
            end

            [~, zz_M_norm] = findpeaks(M_norm(:, jj), 'NPeaks', 8,'SortStr','descend'); % find peaks of correlation function
            min_M_norm(1:length(zz_M_norm),jj) =  sortrows(zz_M_norm,'ascend');% write value into two array
%            
end

 min_M_norm_temp = min_M_norm(1:2,:);
 min_M_norm_temp(:, any(min_M_norm_temp == 0)) = [];%delete if i have one zeros element
 count_M_norm = abs(min_M_norm_temp(1, :) - min_M_norm_temp(2, :)) >= 2;% compare using some condition
 p_M_norm(distance) = sum(count_M_norm);
end
 p_M_level(:,level) = p_M_norm;
end

新增循环后的异常结果 #

0   38
0   38
0   38
0   38
0   35
38  38

问题根源 #

核心问题是循环变量未初始化：在level和distance循环中，M_norm、min_M_norm、p_M_norm这些变量没有在每次循环开始前重置为初始状态，导致上一轮循环的残留数据干扰当前计算。

• 比如当运行level=2时，M_norm还保留着level=1时的计算值，min_M_norm也会累积之前的结果，最终导致后续的count_M_norm和p_M_norm计算逻辑出错。
• 手动修改level值时，每次运行前这些变量都是全新的，没有残留数据，所以结果正确；但嵌套循环时变量会被重复使用，残留数据直接引发异常。

修正方案 #

在每个循环的开头添加变量初始化代码，确保每次循环都从干净的状态开始：

1. 在for level = 1:2内部、for distance = 1:6之前，初始化p_M_norm：

p_M_norm = zeros(6, 1); % 对应distance的6个取值维度

2. 在for distance = 1:6内部、for jj = 1:kol_real之前，初始化M_norm和min_M_norm：

M_norm = zeros(25, kol_real); % 对应dd=1:25和jj=1:kol_real的维度
min_M_norm = zeros(8, kol_real); % 因findpeaks最多取8个峰，行数设为8

修改后，每次循环都会重置相关变量，不会残留上一次的计算数据，结果会和手动修改level值时一致。

内容的提问来源于stack exchange，提问作者Валера Слободянюк