error using the function 'splitapply'

Question

Alberto Acri 2023-11-30

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/2054242-error-using-the-function-splitapply

评论： Alberto Acri 2023-11-30

采纳的回答： Stephen23

在 MATLAB Online 中打开

Hi! I have a problem about using 'splitapply'.

Using the matrix 'matrix_out_98' the code works, but using the matrix 'matrix_out_12' does not work.

Would anyone be able to help me understand why this?

% matrix_out = importdata("matrix_out_12.mat");
matrix_out = importdata("matrix_out_98.mat");
max_matrix_out = max(matrix_out(:,2));
max_matrix_out_r = floor(max_matrix_out*10)/10;
    
%Mention the bins to group data in
j = [0 0.1:0.1:max_matrix_out_r Inf];
%Discretize the data
idx2 = discretize(matrix_out(:,2),j);
%Get the sum of the 2nd column according to the groups
out2 = splitapply(@(x) sum(x), matrix_out(:,2), idx2);

Error:

Error using splitapply (line 111)
For N groups, every integer between 1 and N must occur at least once in the vector of group numbers.

2 个评论
显示无隐藏无

Dyuman Joshi 2023-11-30

移动：Dyuman Joshi 2023-11-30

在 MATLAB Online 中打开

accumarray will be a better fit here.

matrix_out_12 = importdata("matrix_out_12.mat");
matrix_out_98 = importdata("matrix_out_98.mat");
out_12 = fun(matrix_out_12)
out_12 = 22×1
    0.7000
    1.4500
    7.0500
    3.5000
    4.7500
    6.2700
    3.7600
   11.0000
    6.6800
    5.6100
out_98 = fun(matrix_out_98)
out_98 = 17×1
    2.3100
    5.1200
    7.9900
    5.3200
    2.1600
    4.4100
    4.5100
    5.9000
    2.5300
    8.4900
function out = fun(matrix_out);
max_matrix_out = max(matrix_out(:,2));
max_matrix_out_r = floor(max_matrix_out*10)/10;
    
%Mention the bins to group data in
j = [0 0.1:0.1:max_matrix_out_r Inf];
%Discretize the data
idx = discretize(matrix_out(:,2),j);
%Get the sum of the 2nd column according to the groups
out = accumarray(idx, matrix_out(:,2), [], @sum);
end

Dyuman Joshi 2023-11-30

移动：Dyuman Joshi 2023-11-30

splitapply, as the error message states, requires there to be atleast 1 data element in every bin/group in which the data is discretized.

There's no such requirement with accumarray(), where it fills the gap automatically - with the default value of 0, or manually as the value specified (for example NaN)

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Stephen23 2023-11-30

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/2054242-error-using-the-function-splitapply#answer_1362917

编辑：Stephen23 2023-11-30

在 MATLAB Online 中打开

S = load('matrix_out_12.mat') % LOAD is better than IMPORTDATA
S = struct with fields:
    matrix_out: [159×2 double]
max_matrix_out = max(S.matrix_out(:,2));
max_matrix_out_r = floor(max_matrix_out*10)/10; % use ROUND(..,1)
j = [0:0.1:max_matrix_out_r,Inf];
idx2 = discretize(S.matrix_out(:,2),j);

Lets check if all of those groups exist in your data:

ismember(1:max(idx2),idx2) % Nope, some don't:
ans = 1×22 logical array
   1   1   1   1   1   1   1   1   1   1   1   1   1   1   1   1   1   1   0   0   0   1

SPLITAPPLY cannot cope with that: it applies the function to every group from 1 to N, it does not know how to apply your function to data if that data simply does not exist.

Use ACCUMARRAY instead, which in contrast will happily replace undefined outputs with a fill value (by default zero, but you can also set the fill value to whatever you want):

out = accumarray(idx2,S.matrix_out(:,2),[max(idx2),1],@sum)
out = 22×1
    0.7000
    1.4500
    7.0500
    3.5000
    4.7500
    6.2700
    3.7600
   11.0000
    6.6800
    5.6100
out(end-6:end)
ans = 7×1
    1.5700
    3.3400
    5.2100
         0
         0
         0
    2.1200

For example, using NaN as the fill value:

out = accumarray(idx2,S.matrix_out(:,2),[max(idx2),1],@sum, NaN)
out = 22×1
    0.7000
    1.4500
    7.0500
    3.5000
    4.7500
    6.2700
    3.7600
   11.0000
    6.6800
    5.6100
out(end-6:end)
ans = 7×1
    1.5700
    3.3400
    5.2100
       NaN
       NaN
       NaN
    2.1200

7 个评论
显示 5更早的评论隐藏 5更早的评论

Stephen23 2023-11-30

编辑：Stephen23 2023-11-30

在 MATLAB Online 中打开

"'out1' consists of {A×1 double}, should be {A×2 double}"

So you want both columns. Here is an approach using ARRAYFUN:

load("matrix_out_98.mat")
load("idx1_98.mat")
out = arrayfun(@(n)matrix_out(n==idx1,:),1:max(idx1),'Uni',0)
out = 1×17 cell array
  Columns 1 through 12

    {104×2 double}    {34×2 double}    {30×2 double}    {19×2 double}    {5×2 double}    {8×2 double}    {7×2 double}    {8×2 double}    {3×2 double}    {9×2 double}    {7×2 double}    {5×2 double}

  Columns 13 through 17

    {4×2 double}    {13×2 double}    {8×2 double}    {2×2 double}    {[460 1.6400]}

Checking a few cells:

out{5}
ans = 5×2
  410.0000    0.4100
  415.0000    0.4100
  416.0000    0.4400
  418.0000    0.4700
  494.0000    0.4300
out{9}
ans = 3×2
  432.0000    0.8500
  434.0000    0.8100
  483.0000    0.8700