bitset::bitset(const bitset& b)
{
bits.resize(b.size());
for (unsigned int i = 0; i < b.size(); i++)
bits[i] = b[i];
}
/**
* 计算基因型概率
* @param geneBit 基因型的二进制表示，其中1表示大写，0表示小写
* @param r1 左侧重组率
* @param r2 右侧重组率
* @param r 整体重组率，默认为0，通过r1和r2计算
* @param zeta 符合系数，默认为1
* @return 基因型概率
*/
double calcCross(const bitset<GENE_CP_BIT> geneBit,
const double r1, const double r2, const double r=0, const double zeta=1) {
if (geneBit.size()!=GENE_CP_BIT) return 0;
double r12 = r1*r2*zeta;
double r0 = r;
if (r==0) {
r0 = r1+r2-2*r12;
}
// 配子产生的概率，见书105页
switch(geneBit.to_ulong()) {
case 0: //M1QM2 000
case 7: //m1qm2 111
return (1.0-r1-r2+r12);// `=(1-r1)(1-r2);
case 1: //M1Qm2 001
case 6: //m1qM2 110
return (r2-r12);// `=r2(1-r1);
case 4: //M1qm2 100
case 3: //m1QM2 011
return (r1-r12);// `=r1(1-r2);
case 5: //M1qM2 101
case 2: //m1Qm2 010
return r12;// `=r1*r2;
}
return 0;
}
bool covered(bitset<10> digits){
for(int i=0;i<digits.size();i++){
if(digits[i] == false)
return false;
}
return true;
}
// restituisce la posizione del nodo i-esimo in preordine
inline unsigned int preorder_node ( unsigned int n ) const {
unsigned int i = 0;
for ( i = 0; i < tree.size() && n != 0; i++ ) {
n -= tree[i];
}
return i - 1;
}
bool isPrimeUtil(ll N,bitset<100007>&_bitset,vector<int>&primes){
if(N < (int)_bitset.size())
return _bitset.test(N);
for(ll i=0; i<primes.size();i++)
if(N%primes[i]==0)
return false;
return true;
}
void print(bitset<16> &b)
{
int i = 0;
int bsize = b.size();
for (i = 0; i < bsize; ++i)
{
cout << b[i];
}
cout << " :" << "the size of bitset:" << bsize << endl;
}
// restituisco la fine del primo sottoalbero che parte da pos
inline unsigned int first_subtree ( unsigned int pos ) const {
unsigned int z;
unsigned int o;
for ( z = o = 0, pos++; pos < tree.size(); pos++ ) {
z += !tree[pos];
o += tree[pos];
if ( z > o ) break;
}
return pos;
}
bool cart::should_terminate(const vector<const sample*>& samples,
const bitset& splitted_features,
size_t min_num_samples)
{
if (splitted_features.num_bits_set() == splitted_features.size())
{
return true;
}
if (samples.size() < min_num_samples)
{
return true;
}
return false;
}
void logger::setmask(bitset<8>_mask)
{
mask = _mask;
int bits = 0;
for (size_t i = 0; i < _mask.size(); i++) {
if (_mask.test(i)) {
bits |= logmasks[i];
}
}
// cout << "setlogmask(" << hex << bits << dec << ")" << endl;
#ifdef _WIN32
WinLog::SetMask(bits);
#else
setlogmask(bits);
#endif
}
/* Decodes Huffman code to original string.
*
* @param code Huffman code.
* @param root Root of Huffman tree.
* @return Original string.
*/
string decode_huffman(const bitset& code, HuffmanTree* root)
{
string origin;
HuffmanTree* curr_node = root;
for (uint i = 0; i <= code.size(); i++) {
if (curr_node->symbol != 0) {
origin += curr_node->symbol;
curr_node = root;
i--;
} else {
if (code[i] == 0) {
curr_node = curr_node->left;
} else {
curr_node = curr_node->right;
}
}
}
return origin;
}
void cart::find_best_split(vector<const sample*>& samples,
const bitset& splitted_features,
tree_node_split& best_split)
{
/*for each feature f not set in splitted_features
for each splittable value s of feature f
R1 = {samples whose feature f's value is smaller than s}
R2 = {samples whose feature f's value is not smaller than s}
calculate the cost
C = sum{(y1 - c1)^2} + sum{(y2 - c2)^2}
c1 = mean(y) in R1, c2 = mean(y) in R2
find (f,s) with mininum C
*/
//for each feature f
// sort samples by f's value
// scan samples in ascending order of f's value
// for splittable value s
// C = sum(y1 ^ 2) + sum(c1 ^ 2) - 2 * c1 * sum(y1) + sum(y2 ^ 2) + sum(c2 ^ 2) - 2 * c2 * sum(y2)
// C = CL + CR
// c1 = sum(y1) / |CL|
// c2 = sum(y2) / |CR|
// CL = CLY2 + CLC2 - 2 * c1 * CLY
// CR = CRY2 + CRC2 - 2 * c2 * CRY
double min_ds = -1;
best_split.fid = splitted_features.size();
best_split.val = FLT_MIN;
for (size_t fid = 0; fid < splitted_features.size(); fid++)
{
if (splitted_features.test(fid))
{
continue;
}
//sort samples by feature fid
feature_comp fcmp(fid);
std::sort(samples.begin(), samples.end(), fcmp);
double square_cy1, square_cy2, sum_y1, sum_y2, my1, my2, ds1, ds2;
size_t ny1, ny2;
square_cy1 = square_cy2 = sum_y1 = sum_y2 = ny1 = ny2 = my1 = my2 = ds1 = ds2 = 0;
for (vector<const sample*>::iterator it = samples.begin(); it != samples.end(); it++)
{
const sample* sa = *it;
square_cy2 += sa->y * sa->y;
sum_y2 += sa->y;
ny2++;
}
for (vector<const sample*>::iterator it = samples.begin(); it != samples.end(); it++)
{
const sample* sa = *it;
//each sample's value is a split value
square_cy1 += sa->y * sa->y;
ny1++;
sum_y1 += sa->y;
my1 = sum_y1 / ny1;
ds1 = square_cy1 + my1 * my1 * ny1 - 2 * my1 * sum_y1;
square_cy2 -= sa->y * sa->y;
ny2--;
sum_y2 -= sa->y;
my2 = sum_y2 / ny2;
ds2 = square_cy2 + my2 * my2 * ny2 - 2 * my2 * sum_y2;
if (min_ds < 0 || min_ds > ds1 + ds2)
{
min_ds = ds1 + ds2;
best_split.fid = fid;
best_split.val = sa->x[fid];
}
}
}
}
