完成了信息熵计算的程序，解决了bug

.gitignore

@@ -1,3 +1,5 @@
+*.mat
+*.txt
 x64/
 eigen-3.4.0/
 *.debug

CppRun/calc_entropy.cpp

@@ -133,24 +133,25 @@ bool ReadInfoFromMat(const string & filePath, vector<vector<string> >&vvDs, vect
 	return true;
 }
 
+// 将二维索引转成一维的索引
+inline int Get1DIndex(int colNum, int row, int col) {
+	return row * colNum + col;
+}
+
 /* 处理一个知识颗粒 */
-struct EntropyResult { // 存放每个文献对应的结果
-	vector<vector<double> > vvEntropy; // 信息熵
-	vector<vector<double> > vvTransEntropy; // 转置的信息熵
-};
 struct ThreadParam { // 线程参数
 	fs::path matFilePath;
+	fs::path outFilePath;
 	vector<unordered_set<string> >* pvusWord;
-	EntropyResult* pRes;
 };
 void ThreadProcessData(const ThreadParam& param) {
 	const fs::path& matFilePath = param.matFilePath;
-	EntropyResult& res = *param.pRes;
+	const fs::path& outFilePath = param.outFilePath;
 	vector <unordered_set<string> >& vusWord = *param.pvusWord;
 
-	// 存放结果
-	auto& hs = res.vvEntropy;
-	auto& hr = res.vvTransEntropy;
+	// 存放结果，用一维数组存放二维数据
+	vector<double> hs;
+	vector<double> hr;
 
 	vector<vector<string> > vvDs; // 每个知识颗粒的ds矩阵（词汇矩阵）
 	vector<vector<double> > vvFr; // 词汇对应的频率
@@ -162,11 +163,10 @@ void ThreadProcessData(const ThreadParam& param) {
 	// cout << vvDs.size() << '\t' << vvDs[0].size() << endl;
 	const int numLiterature = vusWord.size(); // pubmed 文件中包含的文献数量
 	const int numGroup = vvDs.size(); // ds包含的组数
-	hs.resize(numGroup);
-	hr.resize(numLiterature);
-	for (int i = 0; i < numGroup; ++i) hs[i].resize(numLiterature); // resize会自动初始化
-	for (int i = 0; i < numLiterature; ++i) hr[i].resize(numGroup);
-	for (int groupIdx = 0; groupIdx < vvDs.size(); ++groupIdx) { // 遍历知识颗粒中的每一组
+	hs.resize(numGroup * numLiterature);
+	hr.resize(numLiterature * numGroup);
+
+	for (int groupIdx = 0; groupIdx < numGroup; ++groupIdx) { // 遍历知识颗粒中的每一组
 		vector<string>& vDs = vvDs[groupIdx]; // 这一组ds
 		vector<double>& vFr = vvFr[groupIdx]; // frequency
 		const int numWord = vDs.size(); // 这一组数据中包含的单词数量
@@ -176,27 +176,53 @@ void ThreadProcessData(const ThreadParam& param) {
 			for (int j = 0; j < numWord; ++j) {
 				if (vusWord[i].find(vDs[j]) != vusWord[i].end()) { // 这一组单词中的j索引位置的单词在第i个文献中出现过
 					vX[i][j] = 1;
-					// 对每个知识颗粒每一组数据，计算信息熵
-					hs[groupIdx][i] -= vFr[j] * log2(vFr[j]);
+					if (groupIdx == 1 && i == 2) {
+						// cout << matFilePath.string() << '\t' << j+1 << '\t' << vDs[j] << endl;
+					}
 				}
 			}
 		}
-		// cout << vX[0][0] << endl;
-		
-		
 
+		// 找词汇的最高频率
+		double maxFr = *max_element(vFr.begin(), vFr.end());
+		// 将fr的数值规范化到（0，0.368）之间
+		const double normalMax = 0.368;
+		for (auto& frVal : vFr) frVal = frVal * normalMax / maxFr;
+		maxFr = normalMax;
+		// 对每个知识颗粒每一组数据，计算信息熵
 		for (int i = 0; i < numLiterature; ++i) {
-
-			if (vX[groupIdx][i] == 1) {
-
+			for (int j = 0; j < numWord; ++j) {
+				if (vX[i][j] == 1) {
+					hs[Get1DIndex(numLiterature, groupIdx, i)] -= vFr[j] * log2(vFr[j]);
+				}
 			}
 		}
 
 		// 找最高频词汇所在的索引位置
+		vector<int> vMaxPos;
+		int idx = 0;
+		for_each(vFr.begin(), vFr.end(), [&idx, maxFr, &vMaxPos](double val) {
+			if (val == maxFr) vMaxPos.push_back(idx);
+			idx++;
+		});
+
+		for (int i = 0; i < numLiterature; ++i) {
+			int cumulateX = 0; // 计算在最高频词汇处，x值的累加结果
+			for (int j = 0; j < vMaxPos.size(); ++j) cumulateX += vX[i][vMaxPos[j]];
+			if (cumulateX == vMaxPos.size()) { // 如果频率最高的词汇都出现在了文献中
+				hr[Get1DIndex(numGroup,i, groupIdx)] = 1; // 应该是表示知识颗粒的这一组数据跟这篇文献相关性比较高
+			}
+		}
 	}
+	/* 将结果（hs和hr）写入每个知识颗粒的目录内 */
+	MATFile* pMatFile = matOpen(outFilePath.string().c_str(), "w");
+	SaveMtxDouble(hs.data(), pMatFile, "hs1", numGroup, numLiterature);
+	SaveMtxDouble(hr.data(), pMatFile, "hr1", numLiterature, numGroup);
+	matClose(pMatFile);
 }
 
 /* 程序入口 */
+// 运行例子：CalcEntropy d:\Twirls\runtime\ALS_test abs2class.mat d:\Twirls\runtime\pubmed_files\pubmed-multiplesc-set.mat hx_info.mat 12 word.mat
 void CalcEntropy(int argc, const char** argv) {
 	// argv 
 	// 1. 知识颗粒的父目录名称
@@ -205,12 +231,13 @@ void CalcEntropy(int argc, const char** argv) {
 	// 4. 存放输出结果的mat文件的后缀（每个知识颗粒目录中生成一个结果文件）
 	// 5. 线程数量(可选)
 	if (argc < 5) {
-		cout << "This program should take at least 4 arguments(1.parrent Dir; 2. mat file suffix; 3. pubmed mat file; 4. out mat filename; [5. thread number])!" << endl;
+		cout << "This program should take at least 4 arguments(1.parrent Dir; 2. mat file suffix; 3. pubmed mat file; 4. out mat filename; [5. thread number]; [6. word out mat filepath])!" << endl;
 		return;
 	}
 	clock_t begin, finish;
 	string parrentDir(argv[1]); // 知识颗粒的父目录名称
 	string wordMatSuffix(argv[2]); // 高频词汇矩阵对应的mat文件的后缀名（可以是全文件名，可以是文件名后缀，必须保证唯一）
+	fs::path outFileName(argv[4]);
 	int numThread = 1;
 	if (argc >= 5) numThread = atoi(argv[5]);
 	if (numThread < 1) numThread = 1;
@@ -224,11 +251,13 @@ void CalcEntropy(int argc, const char** argv) {
 		cout << "PubMed Abstract info is null!" << endl;
 		return;
 	}
-	// 将摘要信息分割成一个一个的词汇
+	/* 将摘要信息分割成一个一个的词汇 */
 	begin = clock();
 	unordered_set<char> usWordChars; // 能组成单词的字符，要不要考虑数字？原版matlab是提取了数字的
 	for (int i = 65; i <= 90; i++) usWordChars.insert(char(i)); // A - Z
 	for (int i = 97; i <= 122; i++) usWordChars.insert(char(i)); // a - z
+	for (int i = 48; i <= 57; i++) usWordChars.insert(char(i)); // 0 - 9
+	usWordChars.insert('/'); usWordChars.insert('+'); usWordChars.insert('-');
 	vector<vector<string> > vvWordMtx(vAbstract.size()); // 初始大小为文章的个数
 	vector<unordered_set<string> > vusAbsWord(vAbstract.size()); // 将每篇文章摘要的单词放入hash表
 	for (int i = 0; i < vAbstract.size(); i++) {
@@ -256,6 +285,22 @@ void CalcEntropy(int argc, const char** argv) {
 			vusAbsWord[i].insert(upWord);
 		}
 	}
+	/* 将分割结果写入mat文件 */
+	if (argc >= 6) {
+		MATFile* pMatFile = matOpen(argv[6], "w");
+		mxArray* pCellMtx= mxCreateCellMatrix(1, vvWordMtx.size());
+		for (int i = 0; i < vvWordMtx.size(); ++i) {
+			mxArray* pChildCellMtx = mxCreateCellMatrix(1, vvWordMtx[i].size());
+			for (int j = 0; j < vvWordMtx[i].size(); ++j) {
+				mxArray* mxStr = mxCreateString(vvWordMtx[i][j].c_str());
+				mxSetCell(pChildCellMtx, j, mxStr);
+			}
+			mxSetCell(pCellMtx, i, pChildCellMtx);
+		}
+		matPutVariable(pMatFile, "wd1", pCellMtx);
+		matClose(pMatFile);
+		mxDestroyArray(pCellMtx);
+	}
 	finish = clock();
 	cout << "abstract time: " << (double)(finish - begin) / CLOCKS_PER_SEC << " s" << endl;
 	//auto & vTest = vvWordMtx[0];
@@ -265,20 +310,19 @@ void CalcEntropy(int argc, const char** argv) {
 
 	/* 遍历所有的知识颗粒目录，逐一进行处理 */
 	begin = clock();
-	// ThreadPool thPool(numThread);
-	ThreadPool thPool(24);
+	ThreadPool thPool(numThread);
+	// ThreadPool thPool(24);
 	// 查看知识颗粒数量
 	int numKnowledgeParticle = 0;
 	FOREACH_PARTICLE_START
 		numKnowledgeParticle++;
 	FOREACH_PARTICLE_END
 
-	vector<EntropyResult> vEntropyResult(numKnowledgeParticle); // 存放所有结果
 	// 遍历每个知识颗粒，逐一进行处理
 	for (int round = 0; round < 1; ++round) { // 测试用
 		int i = 0;
 		FOREACH_PARTICLE_START
-			ThreadParam tParam = { file, &vusAbsWord, &vEntropyResult[i] };
+			ThreadParam tParam = { file, childDir / outFileName, &vusAbsWord };
 			thPool.enqueue(ThreadProcessData, tParam);
 			i++;
 		FOREACH_PARTICLE_END
@@ -289,7 +333,7 @@ void CalcEntropy(int argc, const char** argv) {
 	finish = clock();
 
 	cout << "thread pool time: " << (double)(finish - begin) / CLOCKS_PER_SEC << " s" << endl;
-	/* 合并处理结果 */
+
 	//ofstream ofs("test_out.txt");
 	//for (auto& item : vEntropyResult) {
 	//	auto& vvEntropy = item.vvEntropy;