static void rect( int p, mapevalenviron& mapenv, evalcontext* ecntxt )
{
int i, j, p2, p3;
realtyp mid;
realtyp d1, d2;
real2typ xy;
p2 = p + p;
p3 = p2 + p;
for( i = 0; !ecntxt->cycles->error && i < fracsizx; i += p2 )
for( j = p; !ecntxt->cycles->error && j < fracsizy; j += p2 )
{
xy.r2[0] = (realtyp)i/fracsizx;
xy.r2[1] = (realtyp)j/fracsizy;
mid = 0.0;
mid += *adrarar( i , my( j + p) );
mid += *adrarar( i , my( j - p) );
mid += *adrarar( mx( i + p ), j );
mid += *adrarar( mx( i - p ), j );
mid *= 0.25;
mid += normrand( 0.0, stdm( i, j, mapenv, ecntxt ), rndvl );
if ( mid > lmax ) lmax = mid;
else if ( mid < lmin ) lmin = mid;
if ( fracbump && p == 1 )
{
d1 = *adrarar( mx( i + 1 ), j ) - mid;
d2 = *adrarar( i ,my( j + 1 ) ) - mid;
if ( d1 > dmax ) dmax = d1;
else if ( d1 < dmin ) dmin = d1;
if ( d2 > dmax ) dmax = d2;
else if ( d2 < dmin ) dmin = d2;
}
*adrarar( i, j ) = (float)mid;
}
for( i = p; !ecntxt->cycles->error && i < fracsizx; i += p2 )
for( j = 0; !ecntxt->cycles->error && j < fracsizy; j += p2 )
{
xy.r2[0] = (realtyp)i/fracsizx;
xy.r2[1] = (realtyp)j/fracsizy;
mid = 0.0;
mid += *adrarar( i , my( j + p) );
mid += *adrarar( i , my( j - p) );
mid += *adrarar( mx( i + p ), j );
mid += *adrarar( mx( i - p ), j );
mid *= 0.25;
mid += normrand( 0.0, stdm( i, j, mapenv, ecntxt ), rndvl );
if ( mid > lmax ) lmax = mid;
else if ( mid < lmin ) lmin = mid;
if ( fracbump && p == 1 )
{
d1 = *adrarar( mx( i + 1 ), j ) - mid;
d2 = *adrarar( i ,my( j + 1 ) ) - mid;
if ( d1 > dmax ) dmax = d1;
else if ( d1 < dmin ) dmin = d1;
if ( d2 > dmax ) dmax = d2;
else if ( d2 < dmin ) dmin = d2;
}
*adrarar( i, j ) = (float)mid;
}
}
void test_fit(int argc, char *argv[])
{
if (argc < 6) {
printf("usage: %s <n> <s1> <s2> <s3> <s4>\n", argv[0]);
return;
}
int i, n = atoi(argv[1]);
double s1 = atof(argv[2]);
double s2 = atof(argv[3]);
double s3 = atof(argv[4]);
double s4 = atof(argv[5]);
double dx, **X = new_matrix2(n, 4);
for (i = 0; i < n; i++) {
X[i][0] = normrand(0, s1);
X[i][1] = normrand(0, s2);
X[i][2] = normrand(0, s3);
X[i][3] = normrand(0, s4);
dx = norm(X[i], 4);
mult(X[i], X[i], 1/dx, 4);
}
bingham_t B;
bingham_fit(&B, X, n, 4);
free_matrix2(X);
}
void test_mink()
{
double x[10] = {0, 90, 70, 40, 20, 10, 30, 80, 50, 60};
int idx[5];
mink(x, idx, 10, 5);
printf("idx = [ ");
int i;
for (i = 0; i < 5; i++)
printf("%d ", idx[i]);
printf("]\n");
// get timing info
int n = 100000;
double y[n];
for (i = 0; i < n; i++)
y[i] = normrand(0,1);
int yi[n];
double t = get_time_ms();
sort_indices(y, yi, n);
printf("sorted %d numbers in %f ms\n", n, get_time_ms() - t);
t = get_time_ms();
mink(y, yi, n, 100);
printf("got the min %d numbers in %f ms\n", 100, get_time_ms() - t);
}
void test_normrand(int argc, char *argv[])
{
if (argc < 4) {
printf("usage: %s <n> <mu> <sigma>\n", argv[0]);
return;
}
int i, n = atoi(argv[1]);
double mu = atof(argv[2]);
double sigma = atof(argv[3]);
printf("x = [ ");
double x[n];
for (i = 0; i < n; i++) {
x[i] = normrand(mu, sigma);
printf("%f, ", x[i]);
}
printf("]\n");
double sample_mu = sum(x,n) / (double)n;
// set x = (x - sample_mu)^2
for (i = 0; i < n; i++)
x[i] = (x[i] - sample_mu)*(x[i] - sample_mu);
double sample_sigma = sqrt(sum(x,n) / (double)n);
printf("sample mu = %f, sample_sigma = %f\n", sample_mu, sample_sigma);
}
void FELNumerical::generateDistribution(double minpsi, double maxpsi)
{
psi = new double [npart];
double deltpsi = (maxpsi - minpsi)/(npart - 1);
for(unsigned int i = 0; i < npart; i++)
{
psi[i] = minpsi + i*deltpsi;
}
gam = normrand(npart, gamma0, sigmag0);
}
static void diag( int p, mapevalenviron& mapenv, evalcontext* ecntxt )
{
int i, j, p2, p3;
realtyp mid;
p2 = p + p;
p3 = p2 + p;
for( i = p; !ecntxt->cycles->error && i < fracsizx; i += p2 )
for( j = p; !ecntxt->cycles->error && j < fracsizy; j += p2 )
{
mid = 0.0;
mid += *adrarar( mx( i + p ), my( j + p) );
mid += *adrarar( mx( i - p ), my( j + p) );
mid += *adrarar( mx( i + p ), my( j - p) );
mid += *adrarar( mx( i - p ), my( j - p) );
mid *= 0.25;
mid += normrand( 0.0, stdm( i, j, mapenv, ecntxt ), rndvl );
if ( mid > lmax ) lmax = mid;
else if ( mid < lmin ) lmin = mid;
*adrarar( i, j ) = (float)mid;
}
}
void standev(long numtrees, long minwhich, double minsteps,
double *nsteps, double **fsteps, longer seed)
{ /* paired sites tests (KHT or SH) on user-defined trees */
/* used in pars */
long i, j, k;
double wt, sumw, sum, sum2, sd;
double temp;
double **covar, *P, *f, *r;
#define SAMPLES 1000
if (numtrees > maxuser) {
printf("TOO MANY USER-DEFINED TREES");
printf(" test only performed in the first %ld of them\n", (long)maxuser);
} else
if (numtrees == 2) {
fprintf(outfile, "Kishino-Hasegawa-Templeton test\n\n");
fprintf(outfile, "Tree Steps Diff Steps Its S.D.");
fprintf(outfile, " Significantly worse?\n\n");
which = 1;
while (which <= numtrees) {
fprintf(outfile, "%3ld%10.1f", which, nsteps[which - 1]);
if (minwhich == which)
fprintf(outfile, " <------ best\n");
else {
sumw = 0.0;
sum = 0.0;
sum2 = 0.0;
for (i = 0; i < chars; i++) {
if (weight[i] > 0) {
wt = weight[i];
sumw += wt;
temp = (fsteps[which - 1][i] - fsteps[minwhich - 1][i]) / 10.0;
sum += wt*temp;
sum2 += wt * temp * temp;
}
}
temp = sum / sumw;
sd = sqrt(sumw / (sumw - 1.0) * (sum2 - temp * temp));
fprintf(outfile, "%10.1f%12.4f",
(nsteps[which - 1] - minsteps) / 10, sd);
if (sum > 1.95996 * sd)
fprintf(outfile, " Yes\n");
else
fprintf(outfile, " No\n");
}
which++;
}
fprintf(outfile, "\n\n");
} else { /* Shimodaira-Hasegawa test using normal approximation */
if(numtrees > MAXSHIMOTREES){
fprintf(outfile, "Shimodaira-Hasegawa test on first %d of %ld trees\n\n"
, MAXSHIMOTREES, numtrees);
numtrees = MAXSHIMOTREES;
} else {
fprintf(outfile, "Shimodaira-Hasegawa test\n\n");
}
covar = (double **)Malloc(numtrees*sizeof(double *));
for (i = 0; i < numtrees; i++)
covar[i] = (double *)Malloc(numtrees*sizeof(double));
sumw = 0.0;
for (i = 0; i < chars; i++)
sumw += weight[i];
for (i = 0; i < numtrees; i++) { /* compute covariances of trees */
sum = nsteps[i]/(10.0*sumw);
for (j = 0; j <=i; j++) {
sum2 = nsteps[j]/(10.0*sumw);
temp = 0.0;
for (k = 0; k < chars; k++) {
if (weight[k] > 0)
temp = temp + weight[k]*(fsteps[i][k]/10.0-sum)
*(fsteps[j][k]/10.0-sum2);
}
covar[i][j] = temp;
if (i != j)
covar[j][i] = temp;
}
}
for (i = 0; i < numtrees; i++) { /* in-place Cholesky decomposition
of trees x trees covariance matrix */
sum = 0.0;
for (j = 0; j <= i-1; j++)
sum = sum + covar[i][j] * covar[i][j];
if (covar[i][i]-sum <= 0.0)
temp = 0.0;
else
temp = sqrt(covar[i][i] - sum);
covar[i][i] = temp;
for (j = i+1; j < numtrees; j++) {
sum = 0.0;
for (k = 0; k < i; k++)
sum = sum + covar[i][k] * covar[j][k];
if (fabs(temp) < 1.0E-12)
covar[j][i] = 0.0;
else
covar[j][i] = (covar[j][i] - sum)/temp;
}
}
f = (double *)Malloc(numtrees*sizeof(double)); /* resampled sums */
P = (double *)Malloc(numtrees*sizeof(double)); /* vector of P's of trees */
r = (double *)Malloc(numtrees*sizeof(double)); /* store Normal variates */
for (i = 0; i < numtrees; i++)
P[i] = 0.0;
sum2 = nsteps[0]; /* sum2 will be smallest # of steps */
for (i = 1; i < numtrees; i++)
if (sum2 > nsteps[i])
sum2 = nsteps[i];
for (i = 1; i <= SAMPLES; i++) { /* loop over resampled trees */
for (j = 0; j < numtrees; j++) /* draw Normal variates */
r[j] = normrand(seed);
for (j = 0; j < numtrees; j++) { /* compute vectors */
sum = 0.0;
for (k = 0; k <= j; k++)
sum += covar[j][k]*r[k];
f[j] = sum;
}
sum = f[1];
for (j = 1; j < numtrees; j++) /* get min of vector */
if (f[j] < sum)
sum = f[j];
for (j = 0; j < numtrees; j++) /* accumulate P's */
if (nsteps[j]-sum2 <= f[j] - sum)
P[j] += 1.0/SAMPLES;
}
fprintf(outfile, "Tree Steps Diff Steps P value");
fprintf(outfile, " Significantly worse?\n\n");
for (i = 0; i < numtrees; i++) {
fprintf(outfile, "%3ld%10.1f", i+1, nsteps[i]);
if ((minwhich-1) == i)
fprintf(outfile, " <------ best\n");
else {
fprintf(outfile, " %9.1f %10.3f", nsteps[i]-sum2, P[i]);
if (P[i] < 0.05)
fprintf(outfile, " Yes\n");
else
fprintf(outfile, " No\n");
}
}
fprintf(outfile, "\n");
free(P); /* free the variables we Malloc'ed */
free(f);
free(r);
for (i = 0; i < numtrees; i++)
free(covar[i]);
free(covar);
}
} /* standev */
