forked from gkichaev/PAINTOR_V3.0
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Functions_Optimize.h
41 lines (31 loc) · 1.46 KB
/
Functions_Optimize.h
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
//
// Created by Gleb Kichaev on 10/21/15.
//
#ifndef PAINTOR_3_0_FUNCTIONS_OPTIMIZE_H
#define PAINTOR_3_0_FUNCTIONS_OPTIMIZE_H
#include <fstream>
#include <iostream>
#include <vector>
#include <cmath>
#include <Eigen>
#include "nlopt.hpp"
#include <iterator>
#include <string>
#include <iomanip>
#include <algorithm>
#include <random>
#include "Functions_IO.h"
using namespace Eigen;
using namespace std;
using namespace nlopt;
vector<double> GradientFxn(vector<double>& betas, ObjectiveData *in_data);
double ObjectiveFxn(const vector<double> &x, vector<double> &grad, void *data);
void Optimize_Nlopt(vector<double>& x, double lower, double upper, double betaZero, void* in_data);
double dot_prod(vector<double>& vec1, vector<double>& vec2);
vector<double> scalar_product(const double& scalar, vector<double>& vec);
vector<double> Vector_Sum(vector<double>& vec1, vector<double>& vec2);
double Compute_Objective(VectorXd& current_gammas, VectorXd& stacked_probabilites, MatrixXd& stacked_annotations);
void Compute_Gradient(VectorXd& current_gammas, VectorXd& stacked_probabilites, MatrixXd& stacked_annotations, VectorXd& gradient);
void Compute_Hessian(VectorXd& current_gammas, VectorXd& stacked_probabilites, MatrixXd& stacked_annotations, MatrixXd& hessian);
int Gradient_Ascent(VectorXd& current_gammas, VectorXd& new_gammas, VectorXd& stacked_probabilites, MatrixXd& stacked_annotations, double gradient_tolerance, int max_iterations);
#endif //PAINTOR_3_0_FUNCTIONS_OPTIMIZE_H