From f7eebd2376d733cdcfd2f81983c5cef23fe60746 Mon Sep 17 00:00:00 2001
From: carlocamilloni <carlo.camilloni@gmail.com>
Date: Fri, 29 Mar 2019 10:26:27 +0100
Subject: [PATCH] MetaInference: added multiple bias_max bias min entry
---
src/isdb/Metainference.cpp | 97 ++++++++++++++++++++++++++++----------
1 file changed, 73 insertions(+), 24 deletions(-)
diff --git a/src/isdb/Metainference.cpp b/src/isdb/Metainference.cpp
index 27e2c6c00..cbf482469 100644
--- a/src/isdb/Metainference.cpp
+++ b/src/isdb/Metainference.cpp
@@ -314,8 +314,8 @@ void Metainference::registerKeywords(Keywords& keys) {
keys.addOutputComponent("acceptScale", "SCALEDATA", "MC acceptance");
keys.addOutputComponent("weight", "REWEIGHT", "weights of the weighted average");
keys.addOutputComponent("biasDer", "REWEIGHT", "derivatives with respect to the bias");
- keys.addOutputComponent("scale", "default", "scale parameter");
- keys.addOutputComponent("offset", "default", "offset parameter");
+ keys.addOutputComponent("scale", "SCALEDATA", "scale parameter");
+ keys.addOutputComponent("offset", "ADDOFFSET", "offset parameter");
keys.addOutputComponent("ftilde", "GENERIC", "ensemble average estimator");
}
@@ -526,28 +526,43 @@ Metainference::Metainference(const ActionOptions&ao):
vector<double> readsigma;
parseVector("SIGMA0",readsigma);
if((noise_type_!=MGAUSS&&noise_type_!=MOUTLIERS&&noise_type_!=GENERIC)&&readsigma.size()>1)
- error("If you want to use more than one SIGMA you should use NOISETYPE=MGAUSS|MOUTLIERS");
+ error("If you want to use more than one SIGMA you should use NOISETYPE=MGAUSS|MOUTLIERS|GENERIC");
if(noise_type_==MGAUSS||noise_type_==MOUTLIERS||noise_type_==GENERIC) {
- if(readsigma.size()==narg) {
- sigma_.resize(narg);
- sigma_=readsigma;
- } else if(readsigma.size()==1) {
- sigma_.resize(narg,readsigma[0]);
- } else {
- error("SIGMA0 can accept either one single value or as many values as the number of arguments (with NOISETYPE=MGAUSS|MOUTLIERS)");
- }
+ sigma_.resize(readsigma.size());
+ sigma_=readsigma;
} else sigma_.resize(1, readsigma[0]);
- double read_smin_;
- parse("SIGMA_MIN",read_smin_);
- sigma_min_.resize(sigma_.size(),read_smin_);
- double read_smax_;
- parse("SIGMA_MAX",read_smax_);
- sigma_max_.resize(sigma_.size(),read_smax_);
- double read_dsigma_=-1.;
- parse("DSIGMA",read_dsigma_);
- if(read_dsigma_<0) read_dsigma_ = 0.05*(read_smax_ - read_smin_);
- Dsigma_.resize(sigma_.size(),read_dsigma_);
+ vector<double> readsigma_min;
+ parseVector("SIGMA_MIN",readsigma_min);
+ if((noise_type_!=MGAUSS&&noise_type_!=MOUTLIERS&&noise_type_!=GENERIC)&&readsigma_min.size()>1)
+ error("If you want to use more than one SIGMA you should use NOISETYPE=MGAUSS|MOUTLIERS|GENERIC");
+ if(noise_type_==MGAUSS||noise_type_==MOUTLIERS||noise_type_==GENERIC) {
+ sigma_min_.resize(readsigma_min.size());
+ sigma_min_=readsigma_min;
+ } else sigma_min_.resize(1, readsigma_min[0]);
+
+ vector<double> readsigma_max;
+ parseVector("SIGMA_MAX",readsigma_max);
+ if((noise_type_!=MGAUSS&&noise_type_!=MOUTLIERS&&noise_type_!=GENERIC)&&readsigma_max.size()>1)
+ error("If you want to use more than one SIGMA you should use NOISETYPE=MGAUSS|MOUTLIERS|GENERIC");
+ if(noise_type_==MGAUSS||noise_type_==MOUTLIERS||noise_type_==GENERIC) {
+ sigma_max_.resize(readsigma_max.size());
+ sigma_max_=readsigma_max;
+ } else sigma_max_.resize(1, readsigma_max[0]);
+
+ if(sigma_max_.size()!=sigma_min_.size()) error("The number of values for SIGMA_MIN and SIGMA_MAX must be the same");
+
+ vector<double> read_dsigma;
+ parseVector("DSIGMA",read_dsigma);
+ if((noise_type_!=MGAUSS&&noise_type_!=MOUTLIERS&&noise_type_!=GENERIC)&&readsigma_max.size()>1)
+ error("If you want to use more than one SIGMA you should use NOISETYPE=MGAUSS|MOUTLIERS|GENERIC");
+ if(read_dsigma.size()>0) {
+ Dsigma_.resize(read_dsigma.size());
+ Dsigma_=read_dsigma;
+ } else {
+ Dsigma_.resize(sigma_max_.size());
+ for(unsigned i=0; i<sigma_max_.size(); i++) Dsigma_[i] = 0.05*(sigma_max_[i] - sigma_min_[i]);
+ }
// monte carlo stuff
parse("MC_STEPS",MCsteps_);
@@ -564,6 +579,34 @@ Metainference::Metainference(const ActionOptions&ao):
checkRead();
+ // set sigma_bias
+ if(noise_type_==MGAUSS||noise_type_==MOUTLIERS||noise_type_==GENERIC) {
+ if(sigma_.size()==1) {
+ double tmp = sigma_[0];
+ sigma_.resize(narg, tmp);
+ } else if(sigma_.size()>1&&sigma_.size()!=narg) {
+ error("SIGMA0 can accept either one single value or as many values as the number of arguments (with NOISETYPE=MGAUSS|MOUTLIERS|GENERIC)");
+ }
+ if(sigma_min_.size()==1) {
+ double tmp = sigma_min_[0];
+ sigma_min_.resize(narg, tmp);
+ } else if(sigma_min_.size()>1&&sigma_min_.size()!=narg) {
+ error("SIGMA_MIN can accept either one single value or as many values as the number of arguments (with NOISETYPE=MGAUSS|MOUTLIERS|GENERIC)");
+ }
+ if(sigma_max_.size()==1) {
+ double tmp = sigma_max_[0];
+ sigma_max_.resize(narg, tmp);
+ } else if(sigma_max_.size()>1&&sigma_max_.size()!=narg) {
+ error("SIGMA_MAX can accept either one single value or as many values as the number of arguments (with NOISETYPE=MGAUSS|MOUTLIERS|GENERIC)");
+ }
+ if(Dsigma_.size()==1) {
+ double tmp = Dsigma_[0];
+ Dsigma_.resize(narg, tmp);
+ } else if(Dsigma_.size()>1&&Dsigma_.size()!=narg) {
+ error("DSIGMA can accept either one single value or as many values as the number of arguments (with NOISETYPE=MGAUSS|MOUTLIERS|GENERIC)");
+ }
+ }
+
IFile restart_sfile;
restart_sfile.link(*this);
if(getRestart()&&restart_sfile.FileExist(status_file_name_)) {
@@ -683,9 +726,15 @@ Metainference::Metainference(const ActionOptions&ao):
log.printf(" initial data uncertainties");
for(unsigned i=0; i<sigma_.size(); ++i) log.printf(" %f", sigma_[i]);
log.printf("\n");
- log.printf(" minimum data uncertainty %f\n",read_smin_);
- log.printf(" maximum data uncertainty %f\n",read_smax_);
- log.printf(" maximum MC move of data uncertainty %f\n",read_dsigma_);
+ log.printf(" minimum data uncertainties");
+ for(unsigned i=0; i<sigma_.size(); ++i) log.printf(" %f",sigma_min_[i]);
+ log.printf("\n");
+ log.printf(" maximum data uncertainties");
+ for(unsigned i=0; i<sigma_.size(); ++i) log.printf(" %f",sigma_max_[i]);
+ log.printf("\n");
+ log.printf(" maximum MC move of data uncertainties");
+ for(unsigned i=0; i<sigma_.size(); ++i) log.printf(" %f",Dsigma_[i]);
+ log.printf("\n");
log.printf(" temperature of the system %f\n",kbt_);
log.printf(" MC steps %u\n",MCsteps_);
log.printf(" MC stride %u\n",MCstride_);
--
GitLab