這是很久以前寫的一個KDTree,有些時候覺得這個世界真的有點科幻不像是真的#include bits/stdc.h using namespace std; const int featurenum 17; const int featuresize 100; const double inf 1e10; const int K 8; typedef vectordouble Feature; typedef vectorvectordouble Featurelist; Featurelist listfeature; const int Troot 0; vectorint child[featuresize 10]; vectordouble nowfeature; vectorint res; const int Lmax 50; const int kpoint 10; int nodepool featuresize 1; int vst[1005]; int isLeaf[1005]; int L; inline double getDistance(const Feature f1, const Feature f2) { assert(f1.size() featurenum f2.size() featurenum); double ret 0; for (int i 0; i featurenum; i) { ret (f1[i] - f2[i]) * (f1[i] - f2[i]); } return ret; } struct featureNode { //return true; featureNode(int _id) :id(_id) {}; int id; bool operator(const featureNode ano) const { if (getDistance(listfeature[id], nowfeature) getDistance(listfeature[ano.id], nowfeature)) { return false; } return true; } }; priority_queuefeatureNode, vectorfeatureNode Apointq; priority_queuefeatureNode, vectorfeatureNode Fpointq; void generateFeature() { vectordouble tmpFeature; for (int i 0; i featurenum; i) { double tval (rand() % 1000000) / 1000.0; tmpFeature.push_back(tval); } listfeature.push_back(std::move(tmpFeature)); } void randomFeature() { nowfeature.clear(); for (int i 0; i featurenum; i) { double tval (rand() % 1000000) / 1000.0; nowfeature.push_back(tval); } } void addDepth(int u) { if (child[u].size() K) { isLeaf[u] true; return; } for (int i 0; i K; i) { listfeature.push_back(listfeature[child[u][i]]); int tmp child[u][i]; child[u][i] nodepool; child[nodepool - 1].push_back(tmp); } for (int i K; i child[u].size(); i) { double mindis getDistance(listfeature[child[u][0]], listfeature[child[u][i]]); int minid 0; for (int j 1; j K; j) { double tdis getDistance(listfeature[child[u][i]], listfeature[child[u][j]]); if (tdis mindis) { mindis tdis; minid j; } } child[child[u][minid]].push_back(child[u][i]); } child[u].resize(K); for (int i 0; i K; i) { addDepth(child[u][i]); } } void buildKDTree() { for (int i 1; i featuresize; i) { child[0].push_back(i); } addDepth(0); } void traverseTree(featureNode u) { int ui u.id; if (isLeaf[ui]) { L child[ui].size(); for (int i 0; i child[ui].size(); i) { Fpointq.emplace(child[ui][i]); } return; } int minid 0; double mindis getDistance(nowfeature, listfeature[child[ui][0]]); for (int i 1; i child[ui].size(); i) { double tdis getDistance(nowfeature, listfeature[child[ui][i]]); if (tdis mindis) { minid i; mindis tdis; } } for (int i 0; i child[ui].size(); i) { if (i ! minid) { int v child[ui][i]; if (!vst[v]) { Apointq.emplace(v); vst[v] true; } } } featureNode tmp(child[ui][minid]); traverseTree(move(tmp)); } void search() { L 0; while (!Apointq.empty()) Apointq.pop(); while (!Fpointq.empty()) Fpointq.pop(); featureNode tmp(0); vst[0] true; traverseTree(move(tmp)); while (L Lmax !Apointq.empty()) { featureNode fn Apointq.top(); Apointq.pop(); traverseTree(fn); } for (int i 0; i kpoint !Fpointq.empty(); i) { featureNode tmp Fpointq.top(); Fpointq.pop(); res.push_back(tmp.id); } } void init() { vectordouble allzero; for (int i 0; i featurenum; i) { allzero.push_back(0); } listfeature.push_back(move(allzero)); for (int i 0; i featuresize; i) { generateFeature(); } memset(vst, false, sizeof(vst)); memset(isLeaf, false, sizeof(isLeaf)); } int main() { srand(42); init(); buildKDTree(); randomFeature(); search(); for (int i 0; i kpoint; i) { printf(%d , res[i]); } printf(\n); for (int i 0; i featuresize; i) { if (i % 10 0) printf(\n); printf(%4.2f , getDistance(listfeature[i 1], nowfeature)); } }希望有人可以告訴我醫(yī)學界有啥新的突破可以聯(lián)系我微信