💥1 概述
利用图像结构信息是字典学习的难点,针对传统非参数贝叶斯算法对图像结构信息利用不充分,以及算法运行效率低下的问题,该文提出一种二进制数据的贝叶斯非参数聚类算法。
🎉3 参考文献
[1]董道广,芮国胜,田文飚,张洋,刘歌.具有聚类结构相似性的非参数贝叶斯字典学习算法[J].电子与信息学报,2020,42(11):2765-2772.
👨💻4 Matlab代码
clear all; %% add paths to codes and data addpath('Visualization/export_fig/'); addpath('Visualization/'); addpath('BNPBDCA/'); addpath('data/Chemotherapy/') addpath('data/misc/'); outdir='results\Chemotherapy\'; % define output directory, change if necessary if exist(outdir)~=7 mkdir(outdir); end % select chemotherapy drugs interesting_drugs={'Doxorubicin', 'Etoposide', 'Gemcitabine', 'Cisplatin','Docetaxel', 'Methotrexate'} %% Read data data=dlmread('binary_data.txt'); cell_lines=textread('cell_lines.txt','%s'); headers=textread('headers.txt','%s'); data_mut=data(:,14:21);%mutation data Id=sum(data_mut,2)~=0;%find cell lines that have no mutations Im=ismember(headers,interesting_drugs); Im(1:21)=true; data=data(Id,Im); cell_lines=cell_lines(Id); headers=headers(Im); size(data) %% Initialize parameters and hyper parameters fact=0.9; % cooling factor NOIT=500; % no of iterations alpha=1; % beta parameter beta=betaparams(data); % beta parameter gamma=1; % Dirichlet parameter T=1; % initial temperature labels=1:size(data,1); % initial labels c=clust2(data(:,:),NOIT,[outdir 'outfile.txt'],labels,fact,alpha,beta,gamma,T);% cluster data %% Visualization map=flipud(brewermap([],'RdYlBu')); figure;imagesc(data);colormap(map); export_fig(gcf,[outdir 'data'],'-jpg','-q100','-r300','-transparent'); plot_clusters(c,data,[0 0 800 900]);colormap(map);set(gca,'YDir','normal','XTick',1:length(headers),'XTickLabel',headers,'FontSize',12);rotateXLabels(gca,-45); export_fig(gcf,[outdir 'clustered_data'],'-jpg','-q100','-r300','-transparent'); plot_clusters1(c,data,headers,outdir,-1,[0 0 900 250]);
