int NonLinearLSQ::curvefit() {
size_t n(nSize());
size_t p(nParms());
// Initialize the solver function information
_nlsqPointer d = { this };
gsl_multifit_function_fdf mf;
mf.f = &f;
mf.df = &df;
mf.fdf = &fdf;
mf.n = n;
mf.p = p;
mf.params = &d;
const gsl_multifit_fdfsolver_type *T = gsl_multifit_fdfsolver_lmsder;
gsl_multifit_fdfsolver *s = gsl_multifit_fdfsolver_alloc(T, n, p);
_fitParms = guess();
gsl_vector *x = NlsqTogsl(_fitParms);
gsl_matrix *covar = gsl_matrix_alloc(p, p);
gsl_multifit_fdfsolver_set(s, &mf, x);
_nIters = 0;
checkIteration(_nIters, gslToNlsq(s->x), NLVector(p,999.0),
gsl_blas_dnrm2(s->f), GSL_CONTINUE);
do {
_nIters++;
_status = gsl_multifit_fdfsolver_iterate(s);
_fitParms = gslToNlsq(s->x);
gsl_multifit_covar(s->J, 0.0, covar);
_uncert = getUncertainty(covar);
_status = checkIteration(_nIters, _fitParms, _uncert, gsl_blas_dnrm2(s->f),
_status);
if ( _status ) { break; }
if(!doContinue()) { break; }
_status = gsl_multifit_test_delta(s->dx, s->x, absErr(), relErr());
} while ((_status == GSL_CONTINUE) && (_nIters < _maxIters));
// Clean up
gsl_multifit_fdfsolver_free(s);
gsl_matrix_free(covar);
return (_status);
}
void runGenTest(RunParameters &r) {
// Define variables
Vector J, expJ;
std::vector<int> cons;
std::vector<double> weight;
if (r.useGI) {
epsilonP2_ptr=&epsilonP2_GI;
epsilonC_ptr=&epsilonC_GI;
getMaxError_ptr=&getMaxError_GI;
}
else {
epsilonP2_ptr=&epsilonP2;
epsilonC_ptr=&epsilonC;
getMaxError_ptr=&getMaxError;
}
// Get reference sequence from file
FILE *consIn = fopen(r.getConsensusInfile().c_str(),"r");
if (consIn!=NULL) getConsensus(consIn,cons);
else { printf("Error reading input from file %s\n\n",r.getConsensusInfile().c_str()); exit(1); }
fclose(consIn);
if (r.useVerbose) {
printf("Reference sequence: ");
for (int i=0;i<cons.size();i++) printf(" %d",cons[i]);
printf("\n\n");
}
// Retrieve couplings from file
FILE *dataIn=fopen(r.getInfile().c_str(),"r");
if (dataIn!=NULL) getCouplings(dataIn,J);
else { printf("Error reading input from file %s",r.getInfile().c_str()); exit(1); }
fclose(dataIn);
// Resize expJ
for (int i=0;i<J.size();i++) expJ.push_back(std::vector<double>(J[i].size(),0));
for (int i=0;i<J.size();i++) { for (int j=0;j<J[i].size();j++) expJ[i][j] = exp(J[i][j]); }
// Declare 2-point correlations, 3-point correlations, P(k) and magnetisations
bool ThreePoints = (r.p3red || r.p3);
int N = sizetolength(J.size()); // System size
double alpha = 0.01; // Field regularization multiplier
double gamma = 0; // Regularization strength (L2, set below)
if (r.useGamma) {
if (r.gamma==0) gamma=1/(r.sampleB);
else gamma=r.gamma;
}
Vector p(J.size(),std::vector<double>()); // MC magnetisations and 2-point correlations
Vector cc(J.size(),std::vector<double>()); // MC connected 2-point correlations
Vector q(J.size(),std::vector<double>()); // MSA magnetisations and 2-point correlations
Vector qcc(J.size(),std::vector<double>()); // MSA connected 2-point correlations
std::vector<std::vector<std::vector<std::vector<double> > > > p3(N); // MC 3-point correlations
std::vector<std::vector<std::vector<std::vector<double> > > > c3(N); // MC connected 3-point correlations
std::vector<std::vector<std::vector<std::vector<double> > > > q3(N); // MSA 3-point correlations
std::vector<std::vector<std::vector<std::vector<double> > > > qc3(N); // MSA connected 3-point correlations
std::vector<double> pk(N+1,0); // MC mutation probability
std::vector<double> qk(N+1,0); // MSA mutation probability
std::vector<double> absErr(2,0); // Absolute errors on magnetisation and 2-point correlations
for (int i=0;i<J.size();i++) {
cc[i].resize(J[i].size(),0);
p[i].resize(J[i].size(),0);
qcc[i].resize(J[i].size(),0);
q[i].resize(J[i].size(),0);
}
if (ThreePoints) { for (int i=0;i<N;i++) {
p3[i].resize(N);
c3[i].resize(N);
q3[i].resize(N);
qc3[i].resize(N);
for (int j=0;j<N;j++) {
p3[i][j].resize(N);
c3[i][j].resize(N);
q3[i][j].resize(N);
qc3[i][j].resize(N);
for (int k=0;k<N;k++) {
p3[i][j][k].resize(p[i].size()*p[j].size()*p[k].size(),0);
c3[i][j][k].resize(p[i].size()*p[j].size()*p[k].size(),0);
q3[i][j][k].resize(p[i].size()*p[j].size()*p[k].size(),0);
qc3[i][j][k].resize(p[i].size()*p[j].size()*p[k].size(),0);
}
}
} }
// Get sequences from MSA file and compute correlations
FILE *alIn=fopen(r.getInfileAl().c_str(),"r");
FILE *weightIn=fopen(r.getWeights().c_str(),"r");
if (alIn!=NULL){
if (ThreePoints) getAlignment(alIn,weightIn,J,q,q3,qk,cons);
else getAlignment(alIn,weightIn,J,q,qk,cons);
}
else { printf("Error reading input from file %s\n\n",r.getInfileAl().c_str()); exit(1); }
fclose(alIn);
if (weightIn!=NULL) fclose(weightIn);
if (r.useVerbose) printf("Got N=%d, len(h[0])=%d\n",N,(int)J[0].size());
// Get default starting configuration, if nontrivial
std::vector<int> lattice(N);
if (r.useStart) {
FILE *startIn=fopen(r.getStartInfile().c_str(),"r");
for (int i=0;i<N;i++) fscanf(startIn,"%d",&lattice[i]);
}
else { for (int i=0;i<N;i++) lattice[i]=(int) p[i].size(); }
// Prepare to simulate
srand((unsigned)time(0));
// Run MC and get correlations
if (ThreePoints) getErrorGenTest(J, expJ, r.sampleB, r.b, r.runs, p, lattice, pk, p3, cons); // compute errors on P P2 and MAX
else getErrorGenTest(J, expJ, r.sampleB, r.b, r.runs, p, lattice, pk, cons); // compute errors on P P2 and MAX
//Compute connected correlations
double Neff = 0;
double NJeff = 0;
// estimate the threshold for correlations to print out
double meanq = 0;
for (int i=0;i<lattice.size();i++) { for (int a=0;a<p[i].size();a++) {
Neff++;
meanq+=q[i][a];
absErr[0] += (p[i][a] - q[i][a]) * (p[i][a] - q[i][a]);
for (int j=i+1;j<lattice.size();j++) { for (int b=0;b<p[j].size();b++) {
NJeff++;
int idx = index(i,j,lattice.size());
int sab = sindex(a,b,J[i].size(),J[j].size());
absErr[1] += (p[idx][sab] - q[idx][sab]) * (p[idx][sab] - q[idx][sab]);
cc[idx][sab] = p[idx][sab] - (p[i][a] * p[j][b]);
qcc[idx][sab] = q[idx][sab] - (q[i][a] * q[j][b]);
if (ThreePoints) { for (int k=j+1;k<lattice.size();k++) { for (int c=0;c<p[k].size();c++) {
int ijx = idx;
int ikx = index(i,k,lattice.size());
int jkx = index(j,k,lattice.size());
int sac = sindex(a,c,J[i].size(),J[k].size());
int sbc = sindex(b,c,J[j].size(),J[k].size());
int sabc = sindex3(a,b,c,J[i].size(),J[j].size(),J[k].size());
c3[i][j][k][sabc] = p3[i][j][k][sabc] - (p[i][a]*p[jkx][sbc]) - (p[j][b]*p[ikx][sac]) - (p[k][c]*p[ijx][sab]) + (2*(p[i][a]*p[j][b]*p[k][c]));
qc3[i][j][k][sabc] = q3[i][j][k][sabc] - (q[i][a]*q[jkx][sbc]) - (q[j][b]*q[ikx][sac]) - (q[k][c]*q[ijx][sab]) + (2*(q[i][a]*q[j][b]*q[k][c]));
} } }
} }
} }
absErr[0] = sqrt(absErr[0]/Neff);
absErr[1] = sqrt(absErr[1]/NJeff);
meanq=meanq/Neff;
// Print out errors
double maxPrecision=1/(r.sampleB);
double ep1 = epsilonP(q, p, N, maxPrecision, J, gamma, alpha);
double ep2 = (*epsilonP2_ptr)(q, p, N, maxPrecision, J, gamma);
double em = (*getMaxError_ptr)( q, p, maxPrecision, J, gamma, alpha);
printf("\nRelative errors: P %f, P2 %f MAX %f gamma %f\n",ep1,ep2,em,gamma);
printf("Absolute errors: P %f, P2 %f \n\n",absErr[0],absErr[1]);
//Print results for comparison
FILE *mOut = fopen(r.getMOutfile().c_str(),"w");
FILE *pOut = fopen(r.getP2Outfile().c_str(),"w");
FILE *ccOut = fopen(r.getCCOutfile().c_str(),"w");
FILE *pkOut = fopen(r.getPKOutfile().c_str(),"w");
printMagnetisations(mOut, q, p);
double num=0;
printCorrelations(ccOut, qcc, cc,pOut, q, p);
if (ThreePoints){
FILE *p3Out = fopen(r.getP3Outfile().c_str(),"w");
FILE *c3Out = fopen(r.getC3Outfile().c_str(),"w");
if (r.p3red) num=meanq*meanq*meanq;
if (r.p3) num=0;
print3points(c3Out, qc3, c3,p3Out, q3, p3, num);
}
for (int sit=0;sit<N;sit++) fprintf(pkOut,"%d %le %le\n",sit,qk[sit],pk[sit]);
fflush(pkOut);
}
