hungarian源码---matlab实现acc

聚类指标与hungarian算法实现

最新推荐文章于 2023-08-04 22:14:28 发布

原创

最新推荐文章于 2023-08-04 22:14:28 发布 · 653 阅读

4 ·

CC 4.0 BY-SA版权

文章标签：

#聚类 #matlab

文章介绍了聚类的外部和内部评估指标，包括Purity、ARI、NMI、ACC、NCC、Entropy、Compactness和SilhouetteIndex，并提供了Python和Matlab的代码示例。同时，详细阐述了hungarian算法用于解决赋值问题的实现过程，包括代码解析和操作步骤。

详细说明请看：聚类的外部指标(Purity, ARI, NMI, ACC) 和内部指标(NCC,Entropy,Compactness,Silhouette Index)，附代码 (Python 和 Matlab)_聚类指标代码_阿卡西番茄酱的博客-CSDN博客

hungarian代码

function [C,T]=hungarian(A)
%HUNGARIAN Solve the Assignment problem using the Hungarian method.
%
%[C,T]=hungarian(A)
%A - a square cost matrix.
%C - the optimal assignment.
%T - the cost of the optimal assignment.
%s.t. T = trace(A(C,:)) is minimized over all possible assignments.

% Adapted from the FORTRAN IV code in Carpaneto and Toth, "Algorithm 548:
% Solution of the assignment problem [H]", ACM Transactions on
% Mathematical Software, 6(1):104-111, 1980.

% v1.0  96-06-14. Niclas Borlin, niclas@cs.umu.se.
%                 Department of Computing Science, Ume? University,
%                 Sweden. 
%                 All standard disclaimers apply.

% A substantial effort was put into this code. If you use it for a
% publication or otherwise, please include an acknowledgement or at least
% notify me by email. /Niclas

[m,n]=size(A);

if (m~=n)
    error('HUNGARIAN: Cost matrix must be square!');
end

% Save original cost matrix.
orig=A;

% Reduce matrix.
A=hminired(A);

% Do an initial assignment.
[A,C,U]=hminiass(A);

% Repeat while we have unassigned rows.
while (U(n+1))
    % Start with no path, no unchecked zeros, and no unexplored rows.
    LR=zeros(1,n);
    LC=zeros(1,n);
    CH=zeros(1,n);
    RH=[zeros(1,n) -1];
    
    % No labelled columns.
    SLC=[];
    
    % Start path in first unassigned row.
    r=U(n+1);
    % Mark row with end-of-path label.
    LR(r)=-1;
    % Insert row first in labelled row set.
    SLR=r;
    
    % Repeat until we manage to find an assignable zero.
    while (1)
        % If there are free zeros in row r
        if (A(r,n+1)~=0)
            % ...get column of first free zero.
            l=-A(r,n+1);
            
            % If there are more free zeros in row r and row r in not
            % yet marked as unexplored..
            if (A(r,l)~=0 & RH(r)==0)
                % Insert row r first in unexplored list.
                RH(r)=RH(n+1);
                RH(n+1)=r;
                
                % Mark in which column the next unexplored zero in this row
                % is.
                CH(r)=-A(r,l);
            end
        else
            % If all rows are explored..
            if (RH(n+1)<=0)
                % Reduce matrix.
                [A,CH,RH]=hmreduce(A,CH,RH,LC,LR,SLC,SLR);
            end
            
            % Re-start with first unexplored r

最低0.47元/天解锁文章