I wanted to write my own newton fractal generator.. It's using OpenCL... but that's not the problem.. my problem is that atm. only veerryy few pixels are converging.
So to explain what I've done so far:
I've selected a function I wanted to use: f(z)=z^4-1 (for testing purposes)
I've calculated the roots of this function: 1+0î, -1+0î, 0+1î, 0-1î
I've written a OpenCL Host and Kernel:
the kernel uses a struct with 4 doubles: re (real), im (imaginary), r (as abs), phi (as argument, polar angle or how you call it)
computes from resolution, zoom and global_work_id the "type" of the pixel and the intensity - where type is the root the newton method is converging to / whether it's diverging
Here's what I get rendered:
Here is the whole kernel:
#pragma OPENCL EXTENSION cl_khr_fp64 : enable
#define pi 3.14159265359
struct complex {
double im;
double re;
double r;
double phi;
};
struct complex createComplexFromPolar(double _r, double _phi){
struct complex t;
t.r = _r;
t.phi = _phi;
t.re = cos(t.phi)*t.r;
t.im = sin(t.phi)*t.r;
return t;
}
struct complex createComplexFromKarthes(double real, double imag){
struct complex t;
t.re = real;
t.im = imag;
t.phi = atan(imag / real);
t.r = sqrt(t.re*t.re + t.im*t.im);
return t;
}
struct complex recreateComplexFromKarthes(struct complex t){
return t = createComplexFromKarthes(t.re, t.im);
}
struct complex recreateComplexFromPolar(struct complex t){
return t = createComplexFromPolar(t.r, t.phi);
}
struct complex addComplex(const struct complex z, const struct complex c){
struct complex t;
t.re = c.re + z.re;
t.im = c.im + z.im;
return recreateComplexFromKarthes(t);
}
struct complex subComplex(const struct complex z, const struct complex c){
struct complex t;
t.re = z.re - c.re;
t.im = z.im - c.im;
return recreateComplexFromKarthes(t);
}
struct complex addComplexScalar(const struct complex z, const double n){
struct complex t;
t.re = z.re + n;
return recreateComplexFromKarthes(t);
}
struct complex subComplexScalar(const struct complex z, const double n){
struct complex t;
t.re = z.re - n;
return recreateComplexFromKarthes(t);
}
struct complex multComplexScalar(const struct complex z, const double n) {
struct complex t;
t.re = z.re * n;
t.im = z.im * n;
return recreateComplexFromKarthes(t);
}
struct complex multComplex(const struct complex z, const struct complex c) {
return createComplexFromPolar(z.r*c.r, z.phi + c.phi);
}
struct complex powComplex(const struct complex z, int i){
struct complex t = z;
for (int j = 0; j < i; j++){
t = multComplex(t, z);
}
return t;
}
struct complex divComplex(const struct complex z, const struct complex c) {
return createComplexFromPolar(z.r / c.r, z.phi - c.phi);
}
bool compComplex(const struct complex z, const struct complex c, float comp){
struct complex t = subComplex(z, c);
if (fabs(t.re) <= comp && fabs(t.im) <= comp)
return true;
return false;
}
__kernel void newtonFraktal(__global const int* res, __global const int* zoom, __global int* offset, __global const double* param, __global int* result, __global int* resType){
const int x = get_global_id(0) + offset[0];
const int y = get_global_id(1) + offset[1];
const int xRes = res[0];
const int yRes = res[1];
const double a = (x - (xRes / 2)) == 0 ? 0 : (double)(zoom[0] / (x - (double)(xRes / 2)));
const double b = (y - (yRes / 2)) == 0 ? 0 : (double)(zoom[1] / (y - (double)(yRes / 2)));
struct complex z = createComplexFromKarthes(a, b);
struct complex zo = z;
struct complex c = createComplexFromKarthes(param[0], param[1]);
struct complex x1 = createComplexFromKarthes(1,0);
struct complex x2 = createComplexFromKarthes(-1, 0);
struct complex x3 = createComplexFromKarthes(0, 1);
struct complex x4 = createComplexFromKarthes(0, -1);
resType[x + xRes * y] = 3;
int i = 0;
while (i < 30000 && fabs(z.r) < 10000){
z = subComplex(z, divComplex(subComplexScalar(powComplex(z, 4), 1), multComplexScalar(powComplex(z, 3), 4)));
i++;
if (compComplex(z, x1, 0.05)){
resType[x + xRes * y] = 0;
break;
} else if (compComplex(z, x2, 0.05)){
resType[x + xRes * y] = 1;
break;
} else if (compComplex(z, x3, 0.05)){
resType[x + xRes * y] = 2;
break;
}
}
if (fabs(z.r) >= 10000){
resType[x + xRes * y] = 4;
}
result[x + xRes * y] = i;
}
And here is the coloration of the image:
const int xRes = core->getXRes();
const int yRes = core->getYRes();
for (int y = 0; y < fraktal->getHeight(); y++){
for (int x = 0; x < fraktal->getWidth(); x++){
int conDiv = genCL->result[x + y * xRes];
int type = genCL->typeRes[x + y * xRes];
if (type == 0){
//converging to x1
fraktal->setPixel(x, y, 1*conDiv, 1*conDiv, 0, 1);
} else if (type == 1){
//converging to x2
fraktal->setPixel(x, y, 0, 0, 1*conDiv, 1);
} else if (type == 2){
//converging to x3
fraktal->setPixel(x, y, 0, 1*conDiv, 0, 1);
} else if (type == 3){
//diverging and interrupted by loop end
fraktal->setPixel(x, y, 1*conDiv, 0, 0, 1);
} else {
//diverging and interrupted by z.r > 10000
fraktal->setPixel(x, y, 1, 1, 1, 0.1*conDiv);
}
}
}
I had some mistakes in the complex number computations but I check everything today again and again.. I think they should be okay now.. but what else could be the reason that there are just this few start values converging? Did I do something wrong with newton's method?
Thanks for all your help!!
Well somewhat it really helped to run the code as normal C code.. as this makes Debugging easier: so the issue were some coding issues which I have been able to solve now.. for example my pow function was corrupted and when I added or subtracted I forgot to set the imaginary part to the temp complex number .. so here's my final OpenCL kernel:
#pragma OPENCL EXTENSION cl_khr_fp64 : enable
#define pi 3.14159265359
struct complex {
double im;
double re;
double r;
double phi;
};
struct complex createComplexFromPolar(double _r, double _phi){
struct complex t;
t.r = _r;
t.phi = _phi;
t.re = _r*cos(_phi);
t.im = _r*sin(_phi);
return t;
}
struct complex createComplexFromKarthes(double real, double imag){
struct complex t;
t.re = real;
t.im = imag;
t.phi = atan2(imag, real);
t.r = sqrt(t.re*t.re + t.im*t.im);
return t;
}
struct complex recreateComplexFromKarthes(struct complex t){
return createComplexFromKarthes(t.re, t.im);
}
struct complex recreateComplexFromPolar(struct complex t){
return createComplexFromPolar(t.r, t.phi);
}
struct complex addComplex(const struct complex z, const struct complex c){
return createComplexFromKarthes(c.re + z.re, c.im + z.im);
}
struct complex subComplex(const struct complex z, const struct complex c){
return createComplexFromKarthes(z.re - c.re, z.im - c.im);
}
struct complex addComplexScalar(const struct complex z, const double n){
return createComplexFromKarthes(z.re + n,z.im);
}
struct complex subComplexScalar(const struct complex z, const double n){
return createComplexFromKarthes(z.re - n, z.im);
}
struct complex multComplexScalar(const struct complex z, const double n){
return createComplexFromKarthes(z.re * n,z.im * n);
}
struct complex multComplex(const struct complex z, const struct complex c) {
return createComplexFromKarthes(z.re*c.re-z.im*c.im, z.re*c.im+z.im*c.re);
//return createComplexFromPolar(z.r*c.r, z.phi + c.phi);
}
struct complex powComplex(const struct complex z, int i){
struct complex t = z;
for (int j = 0; j < i-1; j++){
t = multComplex(t, z);
}
return t;
}
struct complex divComplex(const struct complex z, const struct complex c) {
return createComplexFromPolar(z.r / c.r, z.phi-c.phi);
}
bool compComplex(const struct complex z, const struct complex c, float comp){
if (fabs(z.re - c.re) <= comp && fabs(z.im - c.im) <= comp)
return true;
return false;
}
__kernel void newtonFraktal(__global const int* res, __global const int* zoom, __global int* offset, __global const double* param, __global int* result, __global int* resType){
const int x = get_global_id(0) + offset[0];
const int y = get_global_id(1) + offset[1];
const int xRes = res[0];
const int yRes = res[1];
const double a = (x - (xRes / 2)) == 0 ? 0 : (double)((x - (double)(xRes / 2)) / zoom[0]);
const double b = (y - (yRes / 2)) == 0 ? 0 : (double)((y - (double)(yRes / 2)) / zoom[1]);
struct complex z = createComplexFromKarthes(a, b);
//struct complex c = createComplexFromKarthes(param[0], param[1]);
struct complex x1 = createComplexFromKarthes(0.7071068, 0.7071068);
struct complex x2 = createComplexFromKarthes(0.7071068, -0.7071068);
struct complex x3 = createComplexFromKarthes(-0.7071068, 0.7071068);
struct complex x4 = createComplexFromKarthes(-0.7071068, -0.7071068);
struct complex f, d;
resType[x + xRes * y] = 11;
int i = 0;
while (i < 6000 && fabs(z.r) < 10000){
f = addComplexScalar(powComplex(z, 4), 1);
d = multComplexScalar(powComplex(z, 3), 3);
z = subComplex(z, divComplex(f, d));
i++;
if (compComplex(z, x1, 0.0000001)){
resType[x + xRes * y] = 0;
break;
} else if (compComplex(z, x2, 0.0000001)){
resType[x + xRes * y] = 1;
break;
} else if (compComplex(z, x3, 0.0000001)){
resType[x + xRes * y] = 2;
break;
} else if (compComplex(z, x4, 0.0000001)){
resType[x + xRes * y] = 3;
break;
}
}
if (fabs(z.r) >= 1000){
resType[x + xRes * y] = 10;
}
result[x + xRes * y] = i;
}
hope it might help someone someday.. :)
Related
I am trying to use this library for some work. In the example, which is given on their website, they use an operator for defining gradient calculation. I want to use a method,i.e., getGradient, instead of an operator. I have tried several ways, including std::bind(), &Rosenbrock::getGradient. None of them works fine. Any idea how this can be done? I don't need a full answer, just hint will be enough.
#include <Eigen/Core>
#include <iostream>
#include <LBFGS.h>
using Eigen::VectorXd;
using namespace LBFGSpp;
class Rosenbrock
{
private:
int n;
public:
Rosenbrock(int n_) : n(n_) {}
double operator()(const VectorXd& x, VectorXd& grad);
double getGradient(const VectorXd& x, VectorXd& grad);
};
double Rosenbrock::operator()(const VectorXd& x, VectorXd& grad){
double fx = 0.0;
for(int i = 0; i < n; i += 2)
{
double t1 = 1.0 - x[i];
double t2 = 10 * (x[i + 1] - x[i] * x[i]);
grad[i + 1] = 20 * t2;
grad[i] = -2.0 * (x[i] * grad[i + 1] + t1);
fx += t1 * t1 + t2 * t2;
}
return fx;
}
double Rosenbrock::getGradient(const VectorXd& x, VectorXd& grad){
double fx = 0.0;
for(int i = 0; i < n; i += 2)
{
double t1 = 1.0 - x[i];
double t2 = 10 * (x[i + 1] - x[i] * x[i]);
grad[i + 1] = 20 * t2;
grad[i] = -2.0 * (x[i] * grad[i + 1] + t1);
fx += t1 * t1 + t2 * t2;
}
return fx;
}
int main(int argc, char** argv){
const int n = 10;
// Set up parameters
LBFGSParam<double> param;
param.epsilon = 1e-6;
param.max_iterations = 100;
// Create solver and function object
LBFGSSolver<double> solver(param);
Rosenbrock fun(n);
// Initial guess
VectorXd x = VectorXd::Zero(n);
double fx;
//int niter = solver.minimize(fun, x, fx);
int niter = solver.minimize(std::bind(Rosenbrock::getGradient, fun, _1, _2), x, fx);
// I want to do something similar to this
std::cout << niter << " iterations" << std::endl;
std::cout << "x = \n" << x.transpose() << std::endl;
std::cout << "f(x) = " << fx << std::endl;
return 0;
}
What about:
struct Bind
{
Rosenbrock & impl;
template <typename X, typename Y> // Template here because I'm lazy writing the full type
double operator () (X x, Y y) { return impl.getGradient(x, y); }
Bind(Rosenbrock & impl) : impl(impl) {}
};
// Then use Bind with your solver:
Bind b(fun);
int niter = solver.minimize(b);
// Example with a template (replace X, Y by the argument signature of the method you are binding)
template <typename T, double (T::*Func)(X, Y)>
struct Bind
{
T & impl;
double operator()(X x, Y y) { return (impl.*Func)(x, y); }
Bind(T & ref) : impl(ref) {}
};
// Using like
Bind<Rosenbrock, &Rosenbrock::getGradient> b(fun);
The above Bind class can be a template. It can be a lambda. You're just redirecting the operator() to the method in the binder's operator ().
I have created a simple cpp project in Visual Studio, but have encountered the follwoing linking error:
LNK2019: unresolved external symbol "public: class std::vector > __thiscall solver::euler(class point_3d (__cdecl*)(class point_3d,double),double,int,class point_3d,double)" (?euler#?$solver#Vpoint_3d####QAE?AV?$vector#Vpoint_3d##V?$allocator#Vpoint_3d###std###std##P6A?AVpoint_3d##V4#N#ZNH0N#Z
I have no idea what may be causing this and would be gratefull for any advice.
Files that constitute the project:
main.cpp
#include "point_3d.h"
#include "solver.h"
point_3d lorentz(point_3d val, double t) {
double x = 10*(val.get_y()-val.get_x());
double y = val.get_x()*(28.0- val.get_z())- val.get_y();
double z = val.get_x()* val.get_y()-(8.0/3)* val.get_z();
return point_3d(x, y, z);
}
int main(){
auto file = std::ofstream("data/results.json");
solver<point_3d> sol;
std::vector<point_3d> result = sol.euler(lorentz, 0.01, 50000, point_3d(1, 1, 1));
}
solver.h
template<typename T>
class solver{
public:
std::vector<T> euler(T (*f)(T val, double s),double h, int steps, T val, double t0 = 0.0);
std::vector<T> back_euler(T (*f)(T val, double s), double h, int steps, T val, double t0 = 0.0,int n=50);
std::vector<T> rk_2(T (*f)(T val, double s), double h, int steps, T val, double t0 = 0.0);
std::vector<T> rk_4(T (*f)(T val, double s), double h, int steps, T val, double t0 = 0.0);
};
solver.cpp
#include "solver.h"
template<typename T>
std::vector<T> solver<T>::euler(T (*f)(T val, double s), double h, int steps, T val, double t0 ) {
std::vector<T>res;
res.resize(steps + 1);
res[0] = val;
double t = t0;
for (int i = 0; i < steps;i++) {
res[i + 1] = res[i] + h * f(res[i], t);
t += h;
}
return res;
}
template<typename T>
std::vector<T> solver<T>::back_euler(T (*f)(T val, double s), double h, int steps, T val, double t0, int n) {
std::vector<T>res;
res.resize(steps + 1);
res[0] = val;
double t = t0;
for (int i = 0; i < steps; i++) {
t += h;
res[i + 1] = res[i];
for (int j = 0; j < n; j++) {
res[i + 1] += res[i] + h * f(res[i + 1],t);
}
}
return res;
}
template<typename T>
std::vector<T> solver<T>::rk_2(T (*f)(T val, double s), double h, int steps, T val, double t0) {
std::vector<T>res;
res.resize(steps + 1);
res[0] = val;
double t = t0;
for (int i = 0; i < steps; i++) {
auto k = h * f(res[i], t) / 2.0;
res[i + 1] = res[i] + h * f(k+res[i],t+h/2.0);
t += h;
}
return res;
}
template<typename T>
std::vector<T> solver<T>::rk_4(T (*f)(T val, double s), double h, int steps, T val, double t0) {
std::vector<T>res;
res.resize(steps + 1);
res[0] = val;
double t = t0;
for (int i = 0; i < steps; i++) {
auto k1 = f(res[i],t);
auto k2 = f(res[i]+h/2.0*k1, t+h/3.0);
auto k3 = f(res[i] + h / 2.0 * k2, t + 2*h / 3.0);
auto k4 = f(res[i] + h * k3, t + h);
res[i + 1] = res[i] + (k1 / 6 + k2 / 3 + k3 / 3 + k4 / 6) * h;
t += h;
}
return res;
}
point_3d.h
class point_3d{
double x;
double y;
double z;
public:
point_3d(double x, double y, double z): x(x), y(y), z(z) {};
double get_x();
double get_y();
double get_z();
point_3d operator+(point_3d other);
point_3d& operator=(point_3d other);
friend point_3d operator*(double k, point_3d p);
};
point_3d.cpp
#include "point_3d.h"
double point_3d::get_x() {
return x;
}
double point_3d::get_y() {
return y;
}
double point_3d::get_z() {
return z;
}
point_3d point_3d::operator+(point_3d other) {
double x = get_x() + other.get_x();
double y = get_y() + other.get_y();
double z = get_z() + other.get_z();
return point_3d(x, y, z);
}
point_3d& point_3d::operator=(point_3d other) {
this->x = other.get_x();
this->y = other.get_y();
this->z = other.get_z();
return *this;
}
point_3d operator*(double k, point_3d p) {
double x = p.get_x();
double y = p.get_y();
double z = p.get_z();
return point_3d(k * x, k * y, k * z);
}
You need to include vector library to use std::vector.
#include <vector>
"Unresolved symbol" means that there is a class or a function declared but not found anywhere in the project (or in the libraries it includes).
I'm translating Python's version of 'page_dewarper' (https://mzucker.github.io/2016/08/15/page-dewarping.html) into C++. I'm going to use dlib, which is a fantastic tool, that helped me in a few optimization problems before. In line 748 of Github repo (https://github.com/mzucker/page_dewarp/blob/master/page_dewarp.py) Matt uses optimize function from Scipy, to find the minimal distance between two vectors. I think, my C++ equivalent should be solve_least_squares_lm() or solve_least_squares(). I'll give a concrete example to analyze.
My data:
a) dstpoints is a vector with OpenCV points - std::vector<cv::Point2f> (I have 162 points in this example, they are not changing),
b) ppts is also std::vector<cv::Point2f> and the same size as dstpoints.
std::vector<cv::Point2f> ppts = project_keypoints(params, input);
It is dependent on:
- dlib::column_vector 'input' is 2*162=324 long and is not changing,
- dlib::column_vector 'params' is 189 long and its values should be changed to get the minimal value of variable 'suma', something like this:
double suma = 0.0;
for (int i=0; i<dstpoints_size; i++)
{
suma += pow(dstpoints[i].x - ppts[i].x, 2);
suma += pow(dstpoints[i].y - ppts[i].y, 2);
}
I'm looking for 'params' vector that will give me the smallest value of 'suma' variable. Least squares algorithm seems to be a good option to solve it: http://dlib.net/dlib/optimization/optimization_least_squares_abstract.h.html#solve_least_squares, but I don't know if it is good for my case.
I think, my problem is that for every different 'params' vector I get different 'ppts' vector, not only single value, and I don't know if solve_least_squares function can match my example.
I must calculate residual for every point. I think, my 'list' from aforementioned link should be something like this:
(ppts[i].x - dstpoints[i].x, ppts[i].y - dstpoints[i].y, ppts[i+1].x - dstpoints[i+1].x, ppts[i+1].y - dstpoints[i+1].y, etc.)
, where 'ppts' vector depends on 'params' vector and then this problem can be solved with least squares algorithm. I don't know how to create data_samples with these assumptions, because it requires dlib::input_vector for every sample, as it is shown in example: http://dlib.net/least_squares_ex.cpp.html.
Am I thinking right?
I'm doing the same thing this days. My solution is writing a Powell Class by myself. It works, but really slowly. The program takes 2 minutes in dewarping linguistics_thesis.jpg.
I don't know what cause the program running so slowly. Maybe because of the algorithm or the code has some extra loop. I'm a Chinese student and my school only have java lessons. So it is normal if you find some extra codes in my codes.
Here is my Powell class.
using namespace std;
using namespace cv;
class MyPowell
{
public:
vector<vector<double>> xi;
vector<double> pcom;
vector<double> xicom;
vector<Point2d> dstpoints;
vector<double> myparams;
vector<double> params;
vector<Point> keypoint_index;
Point2d dst_br;
Point2d dims;
int N;
int itmax;
int ncom;
int iter;
double fret, ftol;
int usingAorB;
MyPowell(vector<Point2d> &dstpoints, vector<double> ¶ms, vector<Point> &keypoint_index);
MyPowell(Point2d &dst_br, vector<double> ¶ms, Point2d & dims);
MyPowell();
double obj(vector<double> ¶ms);
void powell(vector<double> &p, vector<vector<double>> &xi, double ftol, double &fret);
double sign(double a);// , double b);
double sqr(double a);
void linmin(vector<double> &p, vector<double> &xit, int n, double &fret);
void mnbrak(double & ax, double & bx, double & cx,
double & fa, double & fb, double & fc);
double f1dim(double x);
double brent(double ax, double bx, double cx, double & xmin, double tol);
vector<double> usePowell();
void erase(vector<double>& pbar, vector<double> &prr, vector<double> &pr);
};
#include"Powell.h"
MyPowell::MyPowell(vector<Point2d> &dstpoints, vector<double>& params, vector<Point> &keypoint_index)
{
this->dstpoints = dstpoints;
this->myparams = params;
this->keypoint_index = keypoint_index;
N = params.size();
itmax = N * N;
usingAorB = 1;
}
MyPowell::MyPowell(Point2d & dst_br, vector<double>& params, Point2d & dims)
{
this->dst_br = dst_br;
this->myparams.push_back(dims.x);
this->myparams.push_back(dims.y);
this->params = params;
this->dims = dims;
N = 2;
itmax = N * 1000;
usingAorB = 2;
}
MyPowell::MyPowell()
{
usingAorB = 3;
}
double MyPowell::obj(vector<double> &myparams)
{
if (1 == usingAorB)
{
vector<Point2d> ppts = Dewarp::projectKeypoints(keypoint_index, myparams);
double total = 0;
for (int i = 0; i < ppts.size(); i++)
{
double x = dstpoints[i].x - ppts[i].x;
double y = dstpoints[i].y - ppts[i].y;
total += (x * x + y * y);
}
return total;
}
else if(2 == usingAorB)
{
dims.x = myparams[0];
dims.y = myparams[1];
//cout << "dims.x " << dims.x << " dims.y " << dims.y << endl;
vector<Point2d> vdims = { dims };
vector<Point2d> proj_br = Dewarp::projectXY(vdims, params);
double total = 0;
double x = dst_br.x - proj_br[0].x;
double y = dst_br.y - proj_br[0].y;
total += (x * x + y * y);
return total;
}
return 0;
}
void MyPowell::powell(vector<double> &x, vector<vector<double>> &direc, double ftol, double &fval)
{
vector<double> x1;
vector<double> x2;
vector<double> direc1;
int myitmax = 20;
if(N>500)
myitmax = 10;
else if (N > 300)
{
myitmax = 15;
}
double fx2, t, fx, dum, delta;
fval = obj(x);
int bigind;
for (int j = 0; j < N; j++)
{
x1.push_back(x[j]);
}
int iter = 0;
while (true)
{
do
{
do
{
iter += 1;
fx = fval;
bigind = 0;
delta = 0.0;
for (int i = 0; i < N; i++)
{
direc1 = direc[i];
fx2 = fval;
linmin(x, direc1, N, fval);
if (fabs(fx2 - fval) > delta)
{
delta = fabs(fx2 - fval);
bigind = i;
}
}
if (2.0 * fabs(fx - fval) <= ftol * (fabs(fx) + fabs(fval)) + 1e-7)
{
erase(direc1, x2, x1);
return;
}
if (iter >= itmax)
{
cout << "powell exceeding maximum iterations" << endl;
return;
}
if (!x2.empty())
{
x2.clear();
}
for (int j = 0; j < N; j++)
{
x2.push_back(2.0*x[j] - x1[j]);
direc1[j] = x[j] - x1[j];
x1[j] = x[j];
}
myitmax--;
cout << fx2 << endl;
fx2 = obj(x2);
if (myitmax < 0)
return;
} while (fx2 >= fx);
dum = fx - 2 * fval + fx2;
t = 2.0*dum*pow((fx - fval - delta), 2) - delta * pow((fx - fx2), 2);
} while (t >= 0.0);
linmin(x, direc1, N, fval);
direc[bigind] = direc1;
}
}
double MyPowell::sign(double a)//, double b)
{
if (a > 0.0)
{
return 1;
}
else
{
if (a < 0.0)
{
return -1;
}
}
return 0;
}
double MyPowell::sqr(double a)
{
return a * a;
}
void MyPowell::linmin(vector<double>& p, vector<double>& xit, int n, double &fret)
{
double tol = 1e-2;
ncom = n;
pcom = p;
xicom = xit;
double ax = 0.0;
double xx = 1.0;
double bx = 0.0;
double fa, fb, fx, xmin;
mnbrak(ax, xx, bx, fa, fx, fb);
fret = brent(ax, xx, bx, xmin, tol);
for (int i = 0; i < n; i++)
{
xit[i] = (xmin * xit[i]);
p[i] += xit[i];
}
}
void MyPowell::mnbrak(double & ax, double & bx, double & cx,
double & fa, double & fb, double & fc)
{
const double GOLD = 1.618034, GLIMIT = 110.0, TINY = 1e-20;
double val, fw, tmp2, tmp1, w, wlim;
double denom;
fa = f1dim(ax);
fb = f1dim(bx);
if (fb > fa)
{
val = ax;
ax = bx;
bx = val;
val = fb;
fb = fa;
fa = val;
}
cx = bx + GOLD * (bx - ax);
fc = f1dim(cx);
int iter = 0;
while (fb >= fc)
{
tmp1 = (bx - ax) * (fb - fc);
tmp2 = (bx - cx) * (fb - fa);
val = tmp2 - tmp1;
if (fabs(val) < TINY)
{
denom = 2.0*TINY;
}
else
{
denom = 2.0*val;
}
w = bx - ((bx - cx)*tmp2 - (bx - ax)*tmp1) / (denom);
wlim = bx + GLIMIT * (cx - bx);
if ((bx - w) * (w - cx) > 0.0)
{
fw = f1dim(w);
if (fw < fc)
{
ax = bx;
fa = fb;
bx = w;
fb = fw;
return;
}
else if (fw > fb)
{
cx = w;
fc = fw;
return;
}
w = cx + GOLD * (cx - bx);
fw = f1dim(w);
}
else
{
if ((cx - w)*(w - wlim) >= 0.0)
{
fw = f1dim(w);
if (fw < fc)
{
bx = cx;
cx = w;
w = cx + GOLD * (cx - bx);
fb = fc;
fc = fw;
fw = f1dim(w);
}
}
else if ((w - wlim)*(wlim - cx) >= 0.0)
{
w = wlim;
fw = f1dim(w);
}
else
{
w = cx + GOLD * (cx - bx);
fw = f1dim(w);
}
}
ax = bx;
bx = cx;
cx = w;
fa = fb;
fb = fc;
fc = fw;
}
}
double MyPowell::f1dim(double x)
{
vector<double> xt;
for (int j = 0; j < ncom; j++)
{
xt.push_back(pcom[j] + x * xicom[j]);
}
return obj(xt);
}
double MyPowell::brent(double ax, double bx, double cx, double & xmin, double tol = 1.48e-8)
{
const double CGOLD = 0.3819660, ZEPS = 1.0e-4;
int itmax = 500;
double a = MIN(ax, cx);
double b = MAX(ax, cx);
double v = bx;
double w = v, x = v;
double deltax = 0.0;
double fx = f1dim(x);
double fv = fx;
double fw = fx;
double rat = 0, u = 0, fu;
int iter;
int done;
double dx_temp, xmid, tol1, tol2, tmp1, tmp2, p;
for (iter = 0; iter < 500; iter++)
{
xmid = 0.5 * (a + b);
tol1 = tol * fabs(x) + ZEPS;
tol2 = 2.0*tol1;
if (fabs(x - xmid) <= (tol2 - 0.5*(b - a)))
break;
done = -1;
if (fabs(deltax) > tol1)
{
tmp1 = (x - w) * (fx - fv);
tmp2 = (x - v) * (fx - fw);
p = (x - v) * tmp2 - (x - w) * tmp1;
tmp2 = 2.0 * (tmp2 - tmp1);
if (tmp2 > 0.0)
p = -p;
tmp2 = fabs(tmp2);
dx_temp = deltax;
deltax = rat;
if ((p > tmp2 * (a - x)) && (p < tmp2 * (b - x)) &&
fabs(p) < fabs(0.5 * tmp2 * dx_temp))
{
rat = p / tmp2;
u = x + rat;
if ((u - a) < tol2 || (b - u) < tol2)
{
rat = fabs(tol1) * sign(xmid - x);
}
done = 0;
}
}
if(done)
{
if (x >= xmid)
{
deltax = a - x;
}
else
{
deltax = b - x;
}
rat = CGOLD * deltax;
}
if (fabs(rat) >= tol1)
{
u = x + rat;
}
else
{
u = x + fabs(tol1) * sign(rat);
}
fu = f1dim(u);
if (fu > fx)
{
if (u < x)
{
a = u;
}
else
{
b = u;
}
if (fu <= fw || w == x)
{
v = w;
w = u;
fv = fw;
fw = fu;
}
else if (fu <= fv || v == x || v == w)
{
v = u;
fv = fu;
}
}
else
{
if (u >= x)
a = x;
else
b = x;
v = w;
w = x;
x = u;
fv = fw;
fw = fx;
fx = fu;
}
}
if(iter > itmax)
cout << "\n Brent exceed maximum iterations.\n\n";
xmin = x;
return fx;
}
vector<double> MyPowell::usePowell()
{
ftol = 1e-4;
vector<vector<double>> xi;
for (int i = 0; i < N; i++)
{
vector<double> xii;
for (int j = 0; j < N; j++)
{
xii.push_back(0);
}
xii[i]=(1.0);
xi.push_back(xii);
}
double fret = 0;
powell(myparams, xi, ftol, fret);
//for (int i = 0; i < xi.size(); i++)
//{
// double a = obj(xi[i]);
// if (fret > a)
// {
// fret = a;
// myparams = xi[i];
// }
//}
cout << "final result" << fret << endl;
return myparams;
}
void MyPowell::erase(vector<double>& pbar, vector<double>& prr, vector<double>& pr)
{
for (int i = 0; i < pbar.size(); i++)
{
pbar[i] = 0;
}
for (int i = 0; i < prr.size(); i++)
{
prr[i] = 0;
}
for (int i = 0; i < pr.size(); i++)
{
pr[i] = 0;
}
}
I used PRAXIS library, because it doesn't need derivative information and is fast.
I modified the code a little to my needs and now it is faster than original version written in Python.
I'm trying to work out how to write the following:
total = (value * 0.95 ^ 0) + (value * 0.95 ^ 1) + (value * 0.95 ^ 2) ...
or:
x = (y * z ^ 0) + (y * z ^ 1) + (y * z ^ 2) + (y * z ^ 3) ...
This expresses how to calculate x for 4 iterations, but how can I express this to work with a variable number of iterations? Obviously I could create a loop and add the values together, but I'd really like to find a single equation that solves this.
I'm using c++ but I guess this isn't really a language specific problem (sorry I literally don't know where else to ask this question!).
Any ideas?
Thanks,
Chris.
There is no need for a loop here, you "just" need to employ some maths.
Note that you can rewrite that as
y * (z0 + z1 + ... + zn)
Now, the series
z0 + z1 + ... + zn
sums to
(z(n+1) - 1) / (z - 1)
so your equation would be
x = y * (z(n+1) - 1) / (z - 1)
Equation-wise solving, this is a geometric series and can therefore be calculated with
double geometric_series(double y, double z, int N) {
return y * (std::pow(z, N) - 1.0) / (z - 1.0);
}
but the same result can be obtained with some fun C++ metaprogramming: if you know the number of iterations in advanced and you're allowed to use C++17 features and fold expressions you could do as follows
template<std::size_t... N>
double calculate_x(double y, double z, std::index_sequence<N...>) { // [0;N[
auto f = [](double y_p, double z_p, double exp) {
return y_p * std::pow(z_p, exp);
};
return (f(y, z, N) + ...);
}
template <std::size_t N>
auto calculate_x(double y, double z) {
return calculate_x(y, z, std::make_index_sequence<N>{});
}
Alternatively this can also be done with pre-C++17 templates
template <int N>
double calculate_x(double y, double z) {
return calculate_x<N-1>(y, z) + (y * std::pow(z, N - 1));
}
template <>
double calculate_x<0>(double, double) {
return 0;
}
Otherwise a simpler solution would be to just use a loop
double calculate_x_simple(double y, double z, int N) {
double ret = 0.0;
for (int i = 0 ; i < N ; ++i)
ret += y * std::pow(z, i);
return ret;
}
Driver for the code above
int main() {
// x = (y * z ^ 0) + (y * z ^ 1) + (y * z ^ 2) + (y * z ^ 3)
double y = 42.0;
double z = 44.5;
std::cout << (calculate_x<3>(y, z) == calculate_x_simple(y, z, 3)); // 1
}
As you mentioned, it seems reasonable to use a loop. But if you know the amount of iterations at compile time, you could use templates like this:
template <int n>
double foo(double y, double z)
{
return foo<n-1>(y, z) + y * std::pow(z, n);
}
template <>
double foo<-1>(double, double)
{
return 0;
}
With just a little bit of optimisation this will unfold to a single equation.
Example:
#include <iostream>
#include <cmath>
template <int n>
double foo(double y, double z)
{
return foo<n-1>(y, z) + y * std::pow(z, n);
}
template <>
double foo<-1>(double, double)
{
return 0;
}
int main()
{
std::cout << foo<2>(2,3) << std::endl;
}
Output: 26
If a loop would be the only option:
double x = 0;
int n = 5;
for(int exponent = 0; exponent <= n; ++exponent)
x += y*pow(z, exponent);
you can just use math.pow function with a for loop
#include <stdio.h>
#include <math.h>
int main(void) {
int i;
int n = 5;
double y = 0.5;
double z = 0.3;
double answer = 0;
for (i = 0 ; i < n ; i++)
answer += y * pow(z,i);
printf("%f", answer);
return 0;
}
It can be expressed as a sum from n=0 to m. It can be expressed in a single formula, according to wolframalpha.
Don't know if this fulfills your purpose, but you can use recursion(which in real terms is a loop only :) )
int x = evaluate(y, z, count);
int evaluate(y,z, count)
{
if (count <= 0)
return 0;
return (evaluate(y, z, count-1) + y*z^count);
}
Using n as the number of iterations,
#include <cmath>
double foo(double y, double z, int n)
{
double x =0;
for(int i = 0 ; i<n; ++i){
x+=y*std::pow(z,i);
}
return x;
}
Where std::pow is the power function.
When I add some extra formal parameters double tmin=0.0, double tmax=0.0 to the constructor of the Ray in the code below, I always obtain a wrong image with a white top border. These formal parameters currently contribute in no way (i.e. are unused) to the code. So how is it possible to obtain a different image?
System specifications:
OS: Windows 8.1
Compiler: MSVC 2015
Code:
#include "stdafx.h"
#include <math.h>
#include <stdlib.h>
#include <stdio.h>
#include <random>
std::default_random_engine generator(606418532);
std::uniform_real_distribution<double> distribution = std::uniform_real_distribution<double>(0.0, 1.0);
double erand48(unsigned short *x) {
return distribution(generator);
}
#define M_PI 3.14159265358979323846
struct Vector3 {
double x, y, z;
Vector3(double x_ = 0, double y_ = 0, double z_ = 0) { x = x_; y = y_; z = z_; }
Vector3 operator+(const Vector3 &b) const { return Vector3(x + b.x, y + b.y, z + b.z); }
Vector3 operator-(const Vector3 &b) const { return Vector3(x - b.x, y - b.y, z - b.z); }
Vector3 operator*(double b) const { return Vector3(x*b, y*b, z*b); }
Vector3 mult(const Vector3 &b) const { return Vector3(x*b.x, y*b.y, z*b.z); }
Vector3& norm() { return *this = *this * (1 / sqrt(x*x + y*y + z*z)); }
double Dot(const Vector3 &b) const { return x*b.x + y*b.y + z*b.z; } // cross:
Vector3 operator%(Vector3&b) { return Vector3(y*b.z - z*b.y, z*b.x - x*b.z, x*b.y - y*b.x); }
};
//struct Ray { Vector3 o, d; Ray(const Vector3 &o_, const Vector3 &d_, double tmin=0.0, double tmax=0.0) : o(o_), d(d_) {} };
struct Ray { Vector3 o, d; Ray(const Vector3 &o_, const Vector3 &d_) : o(o_), d(d_) {} };
enum Reflection_t { DIFFUSE, SPECULAR, REFRACTIVE };
struct Sphere {
double rad; // radius
Vector3 p, e, f; // position, emission, color
Reflection_t reflection_t; // reflection type (DIFFuse, SPECular, REFRactive)
Sphere(double rad_, Vector3 p_, Vector3 e_, Vector3 f_, Reflection_t reflection_t) :
rad(rad_), p(p_), e(e_), f(f_), reflection_t(reflection_t) {}
double intersect(const Ray &r) const {
Vector3 op = p - r.o;
double t, eps = 1e-4, b = op.Dot(r.d), det = b*b - op.Dot(op) + rad*rad;
if (det<0) return 0; else det = sqrt(det);
return (t = b - det)>eps ? t : ((t = b + det)>eps ? t : 0);
}
};
Sphere spheres[] = {
Sphere(1e5, Vector3(1e5 + 1,40.8,81.6), Vector3(),Vector3(.75,.25,.25),DIFFUSE),//Left
Sphere(1e5, Vector3(-1e5 + 99,40.8,81.6),Vector3(),Vector3(.25,.25,.75),DIFFUSE),//Rght
Sphere(1e5, Vector3(50,40.8, 1e5), Vector3(),Vector3(.75,.75,.75),DIFFUSE),//Back
Sphere(1e5, Vector3(50,40.8,-1e5 + 170), Vector3(),Vector3(), DIFFUSE),//Frnt
Sphere(1e5, Vector3(50, 1e5, 81.6), Vector3(),Vector3(.75,.75,.75),DIFFUSE),//Botm
Sphere(1e5, Vector3(50,-1e5 + 81.6,81.6),Vector3(),Vector3(.75,.75,.75),DIFFUSE),//Top
Sphere(16.5,Vector3(27,16.5,47), Vector3(),Vector3(1,1,1)*.999, SPECULAR),//Mirr
Sphere(16.5,Vector3(73,16.5,78), Vector3(),Vector3(1,1,1)*.999, REFRACTIVE),//Glas
Sphere(600, Vector3(50,681.6 - .27,81.6),Vector3(12,12,12), Vector3(), DIFFUSE) //Lite
};
inline double clamp(double x) { return x<0 ? 0 : x>1 ? 1 : x; }
inline int toInt(double x) { return int(pow(clamp(x), 1 / 2.2) * 255 + .5); }
inline bool intersect(const Ray &r, double &t, int &id) {
double n = sizeof(spheres) / sizeof(Sphere), d, inf = t = 1e20;
for (int i = int(n); i--;) if ((d = spheres[i].intersect(r)) && d<t) { t = d; id = i; }
return t<inf;
}
Vector3 radiance(const Ray &r_, int depth_, unsigned short *Xi) {
double t; // distance to intersection
int id = 0; // id of intersected object
Ray r = r_;
int depth = depth_;
Vector3 cl(0, 0, 0); // accumulated color
Vector3 cf(1, 1, 1); // accumulated reflectance
while (1) {
if (!intersect(r, t, id)) return cl; // if miss, return black
const Sphere &obj = spheres[id]; // the hit object
Vector3 x = r.o + r.d*t, n = (x - obj.p).norm(), nl = n.Dot(r.d)<0 ? n : n*-1, f = obj.f;
double p = f.x>f.y && f.x>f.z ? f.x : f.y>f.z ? f.y : f.z; // max refl
cl = cl + cf.mult(obj.e);
if (++depth>5) if (erand48(Xi)<p) f = f*(1 / p); else return cl; //R.R.
cf = cf.mult(f);
if (obj.reflection_t == DIFFUSE) { // Ideal DIFFUSE reflection
double r1 = 2 * M_PI*erand48(Xi), r2 = erand48(Xi), r2s = sqrt(r2);
Vector3 w = nl, u = ((fabs(w.x)>.1 ? Vector3(0, 1) : Vector3(1)) % w).norm(), v = w%u;
Vector3 d = (u*cos(r1)*r2s + v*sin(r1)*r2s + w*sqrt(1 - r2)).norm();
r = Ray(x, d);
continue;
}
else if (obj.reflection_t == SPECULAR) {
r = Ray(x, r.d - n * 2 * n.Dot(r.d));
continue;
}
Ray reflRay(x, r.d - n * 2 * n.Dot(r.d));
bool into = n.Dot(nl)>0;
double nc = 1, nt = 1.5, nnt = into ? nc / nt : nt / nc, ddn = r.d.Dot(nl), cos2t;
if ((cos2t = 1 - nnt*nnt*(1 - ddn*ddn))<0) {
r = reflRay;
continue;
}
Vector3 tdir = (r.d*nnt - n*((into ? 1 : -1)*(ddn*nnt + sqrt(cos2t)))).norm();
double a = nt - nc, b = nt + nc, R0 = a*a / (b*b), c = 1 - (into ? -ddn : tdir.Dot(n));
double Re = R0 + (1 - R0)*c*c*c*c*c, Tr = 1 - Re, P = .25 + .5*Re, RP = Re / P, TP = Tr / (1 - P);
if (erand48(Xi)<P) {
cf = cf*RP;
r = reflRay;
}
else {
cf = cf*TP;
r = Ray(x, tdir);
}
continue;
}
}
int main(int argc, char *argv[]) {
int w = 512, h = 384, samps = argc == 2 ? atoi(argv[1]) / 4 : 1; // # samples
Ray cam(Vector3(50, 52, 295.6), Vector3(0, -0.042612, -1).norm()); // cam pos, dir
Vector3 cx = Vector3(w*.5135 / h), cy = (cx%cam.d).norm()*.5135, r, *c = new Vector3[w*h];
#pragma omp parallel for schedule(dynamic, 1) private(r) // OpenMP
for (int y = 0; y<h; y++) { // Loop over image rows
fprintf(stderr, "\rRendering (%d spp) %5.2f%%", samps * 4, 100.*y / (h - 1));
for (unsigned short x = 0, Xi[3] = { 0,0,y*y*y }; x<w; x++) // Loop cols
for (int sy = 0, i = (h - y - 1)*w + x; sy<2; sy++) // 2x2 subpixel rows
for (int sx = 0; sx<2; sx++, r = Vector3()) { // 2x2 subpixel cols
for (int s = 0; s<samps; s++) {
double r1 = 2 * erand48(Xi), dx = r1<1 ? sqrt(r1) - 1 : 1 - sqrt(2 - r1);
double r2 = 2 * erand48(Xi), dy = r2<1 ? sqrt(r2) - 1 : 1 - sqrt(2 - r2);
Vector3 d = cx*(((sx + .5 + dx) / 2 + x) / w - .5) +
cy*(((sy + .5 + dy) / 2 + y) / h - .5) + cam.d;
r = r + radiance(Ray(cam.o + d * 140, d.norm()), 0, Xi)*(1. / samps);
} // Camera rays are pushed ^^^^^ forward to start in interior
c[i] = c[i] + Vector3(clamp(r.x), clamp(r.y), clamp(r.z))*.25;
}
}
FILE *fp;
fopen_s(&fp, "image.ppm", "w"); // Write image to PPM file.
fprintf(fp, "P3\n%d %d\n%d\n", w, h, 255);
for (int i = 0; i<w*h; i++)
fprintf(fp, "%d %d %d ", toInt(c[i].x), toInt(c[i].y), toInt(c[i].z));
}
First Ray structure:
struct Ray { Vector3 o, d; Ray(const Vector3 &o_, const Vector3 &d_) : o(o_), d(d_) {} };
Results in:
Second Ray structure:
struct Ray { Vector3 o, d; Ray(const Vector3 &o_, const Vector3 &d_, double tmin=0.0, double tmax=0.0) : o(o_), d(d_) {} };
Results in:
The last image has a noticeable white top border which is not present in the first image.
Edit:
I used
size_t n = sizeof(spheres) / sizeof(Sphere);
Now I obtain the same images, but I also checked if the original int(n) could differ from 9 which is never the case.
Ok this is from the Debug build, which is different from the Release build.
Sounds like a memory error, looking quickly at your code I'm sceptical of this line:
for (int i = int(n); i--;) if ((d = spheres[i].intersect(r)) && d<t)
I suspect accessing sphere[i] is out of bounds, perhaps you should try sphere[i-1]. You could also try compiling your code with a compiler that adds extra code for debugging/sanitising/checking memory addresses.