double ran_normal()
{
static int iset = 0;
static double gset;
double fac, rsq, v1, v2;
if (iset == 0) {
do {
v1 = 2.0 * ran_uniform() - 1.0;
v2 = 2.0 * ran_uniform() - 1.0;
rsq = v1 * v1 + v2 * v2;
} while (rsq >= 1.0 || rsq == 0.0);
fac = sqrt(-2.0 * log(rsq) / rsq);
gset = v1 * fac;
iset = 1;
return v2 * fac;
} else {
iset = 0;
return gset;
}
}
void logistic_regression::fit(std::string train_dir,double stop){
srand(time(0));
File_Control fc(train_dir);
stop *= (1000*1000);
long pos = 0;
long neg = 0;
no_train = 0;
update = true;
//训练第几遍文件
int indx = 0;
for(int i=0;i<iter;i++){
std::cout<<"训练第"<<i<<"遍文件"<<std::endl;
while(fc.has_next() && indx <= stop){
std::string next = fc.next();
if(next.size()>3){
//subsampling
double ran = ran_uniform();
std::pair<int,Sparse_Vector> result = sparse_vector_form(next);
if(result.first==1 || (ran<1)){
if (result.first==1)
pos ++ ;
else if (result.first==0)
neg ++ ;
one_round(result.first,result.second);
}
indx ++ ;
if (indx%(verbose*1000)==0)
std::cout<<"indx "<<std::setw(5)<<(double)indx/(1000*1000)<<"M |w "<<w.size()<<" |z "<<z.size() \
<<" |g "<<g.size()<<" |no_train "<<no_train<<" |n "<<n.size() \
<<" |pos "<<pos<<" |neg "<<neg<<std::endl;
}
}
if(indx>=stop){
indx = 0;
}
indx = 0;
fc.restart();
}
}
const int poisson_random_number(const double lambda) //Poisson distribution
{
int k=0;
const int max_k = 1000;
double p = ran_uniform();
double P = exp(-lambda);
double sum=P;
if (sum>=p) return 0;
for (k=1; k<max_k; ++k) {
P*=lambda/(double)k;
sum+=P;
if (sum>=p) break;
}
return k;
}
