initial codes

init_cnn.cpp

@@ -0,0 +1,224 @@
+#include <cstring>
+#include <cstdio>
+#include <string>
+#include <cstdlib>
+#include <cmath>
+#include <algorithm>
+#include <vector>
+
+using namespace std;
+
+string path = "./data/FB60K/";
+
+vector<int> Nalist, FNalist;
+int NA = -1;
+const float pi = 3.141592653589793238462643383;
+int rate = 3;
+
+float rand(float min, float max) {
+	return min + (max - min) * rand() / (RAND_MAX + 1.0);
+}
+
+float normal(float x, float miu,float sigma) {
+	return 1.0/sqrt(2*pi)/sigma*exp(-1*(x-miu)*(x-miu)/(2*sigma*sigma));
+}
+
+float randn(float miu,float sigma, float min ,float max) {
+	float x, y, dScope;
+	do {
+		x = rand(min,max);
+		y = normal(x,miu,sigma);
+		dScope=rand(0.0,normal(miu,miu,sigma));
+	} while (dScope > y);
+	return x;
+}
+
+int word_size;
+int entity_size;
+int dimension;
+int PositionLimit = 30;
+int LenLimit = 100;
+float *word_embeddings;
+
+struct Tip {
+	int h;
+	int t;
+	int r;
+	int tot;
+	int *lists;
+};
+
+Tip *tipList;
+int tipTotal;
+int midTotal;
+int relationTotal = 0;
+int sentenceTotal;
+int instanceTot = 0;
+int *sentence, *posH, *posT, *bags_train;
+
+int getPosition(int position) {
+	if (position < -PositionLimit) return 0;
+	if (position > PositionLimit) return 2 * PositionLimit;
+	return position + PositionLimit;
+}
+
+extern "C"
+void setNA(int con) {
+	NA = con;
+}
+
+extern "C"
+void setRate(int con) {
+	rate = con;
+}
+
+extern "C"
+void readWordVec() {
+
+	FILE *fin1 = fopen((path + "entity2id.txt").c_str(), "r");
+	FILE *fin2 = fopen((path + "vec.txt").c_str(), "r");
+	fscanf(fin2, "%d%d\n", &word_size, &dimension);
+	fscanf(fin1, "%d", &entity_size);
+
+	char buffer[100];
+	word_size += 2 + entity_size;
+	word_embeddings = (float *)calloc(word_size * dimension, sizeof(float));
+	int last = 0;
+	for (int i = entity_size * dimension; i > 0 ; i--)
+		word_embeddings[last++] = randn(0, 1.0 / dimension, -6 / sqrt(dimension), 6 / sqrt(dimension));
+	last += 2 * dimension;
+	for (int i = entity_size + 2; i < word_size; i++) {
+		fscanf(fin2, "%s", buffer);
+		for (int i = 0; i < dimension; i++)
+			fscanf(fin2, "%f", &word_embeddings[last++]);
+	}
+
+	fclose(fin2);
+	fclose(fin1);
+}
+
+extern "C"
+void getWordVec(float *con) {
+	for (int i = word_size * dimension - 1; i >= 0; i--)
+		con[i] = word_embeddings[i];
+}
+
+int n = 0;
+int lastNA = 0;
+int lastFNA = 0;
+extern "C"
+int batch_iter(int *x_batch, int *p_h_batch, int *p_t_batch, int *y_batch, int *r_batch, float *r_n_batch, int *h_batch, int *t_batch) {
+	 if (rate == 0 || Nalist.size() == 0) {
+	 	n = FNalist[lastFNA];
+	 	r_n_batch[0] = 1;
+	 	lastFNA = (lastFNA + 1) % (FNalist.size());
+	 } else if (rand() % rate != 0) {
+	 	n = Nalist[lastNA];
+	 	r_n_batch[0] = 0;
+	 	lastNA = (lastNA + 1) % (Nalist.size());
+	 } else {
+	 	n = FNalist[lastFNA];
+	 	r_n_batch[0] = 1;
+	 	lastFNA = (lastFNA + 1) % (FNalist.size());
+	 }
+	int instance = tipList[n].tot;
+	int last = 0;
+	for (int i = 0; i < instance; i++) {
+		int j = tipList[n].lists[i];
+		int last1 = j * LenLimit;
+		for (int k = 0; k < LenLimit; k++) {
+			x_batch[last] = sentence[last1];
+			p_h_batch[last] = posH[last1];
+			p_t_batch[last] = posT[last1];
+			last++;
+			last1++;
+		}
+	}
+	r_batch[0] = tipList[n].r;
+	for (int j = 0; j < relationTotal; j++)
+		y_batch[j] = 0;
+	y_batch[tipList[n].r] = 1;
+	h_batch[0] = tipList[n].h;
+	t_batch[0] = tipList[n].t;
+	return instance;
+}
+
+extern "C"
+int getTipTotal() {
+	return tipTotal;
+}
+
+extern "C"
+int getLenLimit() {
+	return LenLimit;
+}
+
+extern "C"
+int getRelationTotal() {
+	return relationTotal;
+}
+
+extern "C"
+int getWordTotal() {
+	return word_size;
+}
+
+extern "C"
+int getPositionLimit() {
+	return PositionLimit;
+}
+
+extern "C"
+int getWordDimension() {
+	return dimension;
+}
+
+extern "C"
+int getInstanceTot() {
+	return instanceTot;
+}
+
+extern "C"
+void readFromFile() {
+	FILE *f = fopen((path + "corpus2id.txt").c_str(), "r");
+	fscanf(f, "%d\n", &tipTotal);
+	fscanf(f, "%d\n", &sentenceTotal);
+	tipList = (Tip *)calloc(tipTotal, sizeof(Tip));
+	sentence = (int *)calloc(sentenceTotal * LenLimit, sizeof(int));
+	posH = (int *)calloc(sentenceTotal * LenLimit, sizeof(int));
+	posT = (int *)calloc(sentenceTotal * LenLimit, sizeof(int));
+	bags_train = (int *)calloc(sentenceTotal, sizeof(int));
+	int h, t, post, posh, r, len, tip;
+	for (int i = 0; i < sentenceTotal; i++) {
+		fscanf(f, "%d%d%d%d%d%d%d",&h, &t, &posh, &post, &r, &tip, &len);
+		int last = i * LenLimit;
+		for (int j = 0; j < len; j++) {
+			fscanf(f, "%d", &sentence[last + j]);
+			posH[last + j] = getPosition(j - posh);
+			posT[last + j] = getPosition(j - post);
+		}
+		bags_train[i] = tip;
+		tipList[tip].tot++;
+		tipList[tip].h = h;
+		tipList[tip].r = r;
+		tipList[tip].t = t;
+		if (r + 1 > relationTotal)
+			relationTotal = r + 1;
+		if (tipList[tip].tot > instanceTot)
+			instanceTot = tipList[tip].tot;
+	}
+	fclose(f);
+	for (int i = 0; i < tipTotal; i++) {
+		tipList[i].lists = new int[tipList[i].tot];
+		tipList[i].tot = 0;
+		if (tipList[i].r == NA) Nalist.push_back(i); else FNalist.push_back(i);
+	}
+	for (int i = 0; i < sentenceTotal; i++)
+		tipList[bags_train[i]].lists[tipList[bags_train[i]].tot++] = i;
+}
+
+int main() {
+	readWordVec();
+	readFromFile();
+	return 0;
+}