Reputation: 26037
I'm trying to play with some code..keep getting an compile error RND not declared in scope
I found a part of the code that defined it if it ran on linux and if defined it on windows thus ignoring Mac users(no biggie, I would ignore them too!). I removed that part of the code and defined it using the linux settings(since I figured my Mac is more closer to linux than windows), but then I get the same error but for seed. The odd thing is those seed errors are at the same spot at the RND error was. So my question is what the heck is RND/Seed? My searches found them specific to VB but not sure if its useful since I'm using C++.
Here's an offensive code snipped(viewers discretion is advised):
mi = (int)(round(RND*(dimc-1)));
Any tips/suggestions would be great. I'm just starting to learn about c++ so I maybe missing something very simple.
Here's the entire code(stole it from here http://cg.iit.bme.hu/~zsolnai/gfx/genetic/ ):
// a fast genetic algorithm for the 0-1 knapsack problem
// by karoly zsolnai - [email protected]
// test case: 1000 items, 50 knapsack size
//
// compilation by: g++ genetic.cpp -O3 -ffast-math -fopenmp
#include <math.h>
#include <time.h>
#include <algorithm>
#include <vector>
#include <fstream>
#include <limits.h>
#define RND ((double)rand_r(&seed)/RAND_MAX) // reentrant uniform rnd
using namespace std;
struct chromo {
chromo(int dimc) { items = new bool[dimc]; }
~chromo() { items = NULL; }
void mutate(const int dimc, const int count) {
int mi;
for(int i=0;i<count;i++) {
mi = (int)(round(RND*(dimc-1)));
items[mi] = !items[mi];
}
}
bool* items;
int f;
};
int fitness(bool*& x, const int dimc, const vector<int>& v, const vector<int>& w, const int limit) {
int fit = 0, wsum = 0;
for(int i=0;i<dimc;i++) {
wsum += x[i]*w[i];
fit += x[i]*v[i];
}
if(wsum>limit) fit -= 7*(wsum-limit); // penalty for invalid solutions
return fit;
}
void crossover1p(const chromo& c1, const chromo& c2, const chromo& c3, const int dimc, const int cp) {
for(int i=0;i<dimc;i++) {
if(i<cp) { c3.items[i] = c1.items[i]; }
else { c3.items[i] = c2.items[i]; }
}
}
void crossover1p_b(const chromo &c1, const chromo &c2, const chromo &c3, int dimc, int cp) {
for(int i=0;i<dimc;i++) {
if(i>=cp) { c3.items[i] = c1.items[i]; }
else { c3.items[i] = c2.items[i]; }
}
}
void crossoverrand(const chromo &c1, const chromo &c2, const chromo &c3, const int dimc) {
for(int i=0;i<dimc;i++) {
if(round(RND)) { c3.items[i] = c1.items[i]; }
else { c3.items[i] = c2.items[i]; }
}
}
void crossoverarit(const chromo &c1, const chromo &c2, const chromo &c3, int dimc) {
for(int i=0;i<dimc;i++) {
c3.items[i] = (c1.items[i]^c2.items[i]);
}
}
bool cfit(const chromo &c1,const chromo &c2) { return c1.f > c2.f; }
bool cmpfun(const std::pair<int,double> &r1, const std::pair<int,double> &r2) { return r1.second > r2.second; }
int coin(const double crp) { // a cointoss
if(RND<crp) return 1; // crossover
else return 0; // mutation
}
// initializes the chromosomes with the results of a greedy algorithm
void initpopg(bool**& c, const std::vector<int> &w, const std::vector<int> &v, const int dimw, const int limit, const int pop) {
std::vector<std::pair<int,double> > rvals(dimw);
std::vector<int> index(dimw,0);
for(int i=0;i<dimw;i++) {
rvals.push_back(std::pair<int,double>(std::make_pair(i,(double)v[i]/(double)w[i])));
}
std::sort(rvals.begin(),rvals.end(),cmpfun);
int currentw = 0, k;
for(int i=0;i<dimw;i++) {
k = rvals[i].first;
if(currentw + w[k] <= limit) { // greedy fill
currentw += w[k];
index[k] = 1;
}
}
for(int i=0;i<pop;i++) {
for(int j=0;j<dimw;j++) {
c[i][j] = index[j];
}
}
}
int main() {
printf("\n");
srand(time(NULL));
vector<int> w, v; // items weights and values
int info=0;
FILE *f = fopen("1000_weights.txt","r");
FILE *f2 = fopen("1000_values.txt","r");
while(!feof(f) || !feof(f2) ) {
fscanf(f," %d ",&info);
w.push_back(info);
info=0;
fscanf(f2," %d ",&info);
v.push_back(info);
} // omitted fclose(f1) and fclose(f2) on purpose
const int limit = 50; // knapsack weight limit
const int pop = 250; // chromosome population size
const int gens = INT_MAX; // maximum number of generations
const int disc = (int)(ceil(pop*0.8)); // chromosomes discarded via elitism
const int dimw = w.size();
int best = 0, ind = 0, ind2 = 0; // a few helpers for the main()
int parc = 0; // parent index for crossover
double avg = 0, crp = 0.35; // crossover probability
vector<chromo> ch(pop,chromo(dimw));
bool **c = new bool*[pop];
for(int i=0;i<pop;i++) c[i] = new bool[dimw];
clock_t start = clock();
printf("Initializing population with a greedy algorithm...");
initpopg(c,w,v,dimw,limit,pop);
printf("done!");
for(int i=0;i<pop;i++) {
ch[i].items = c[i];
ch[i].f = fitness(ch[i].items, dimw ,v, w, limit);
}
printf("\n\n");
for(int p=0;p<gens;p++) {
std::sort(ch.begin(), ch.end(), cfit);
#pragma omp parallel for shared(ch)
for(int i=0;i<pop;i++) {
if(i>pop-disc) { // elitism - only processes the discarded chromosomes
if(coin(crp)==1) { // crossover section
ind = parc+round(10*RND); // choosing parents for crossover
ind2 = parc+1+round(10*RND);
// choose a crossover strategy here
crossover1p(ch[ind%pop],ch[ind2%pop],ch[i],dimw,round(RND*(dimw-1)));
// crossoverrand(ch[ind],ch[ind2],ch[i],dimw);
// crossoverarit(ch[0],ch[1],ch[i],dimw);
ch[i].f = fitness(ch[i].items, dimw ,v, w, limit);
parc += 1;
}
else { // mutation section
ch[i].mutate(dimw,1);
ch[i].f = fitness(ch[i].items, dimw ,v, w, limit);
}
}
avg += ch[i].f;
if(ch[i].f>best) best=ch[i].f;
}
parc = 0;
if(p%5==0) {
printf("\n#%d\t",p);
printf("best fitness: %d \t",best);
printf("avg fitness: %f",avg/pop);
if(best == 675) goto end; // psst...don't tell anyone
}
best = avg = 0;
}
end:
printf("\n\n");
clock_t end = clock();
double t = (double)(end-start)/CLOCKS_PER_SEC;
printf("\nCompletion time: %fs.\n",t);
return 0;
}
Upvotes: 2
Views: 951
Reputation: 104080
The problem is you've inexpertly cut apart the code you've got:
#if defined(__linux) || defined(__linux__)
unsigned int seed = time(NULL);
#define RND ((double)rand_r(&seed)/RAND_MAX) // reentrant uniform rnd
#endif
#if defined(WIN32) || defined(_WIN32) || defined(__WIN32__)
#define RND ((double)rand()/RAND_MAX) // uniform rnd
#endif
This defines the seed
variable based on the current time for Linux systems; perhaps the Windows systems do not need a seed?
In any event, if you include both lines from the if defined (__linux) ...
branch, instead of only one line, it should work without trouble on your OS X system.
Upvotes: 3