Line data Source code
1 : /* +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
2 : Copyright (c) 2016-2023 The plumed team
3 : (see the PEOPLE file at the root of the distribution for a list of names)
4 :
5 : See http://www.plumed.org for more information.
6 :
7 : This file is part of plumed, version 2.
8 :
9 : plumed is free software: you can redistribute it and/or modify
10 : it under the terms of the GNU Lesser General Public License as published by
11 : the Free Software Foundation, either version 3 of the License, or
12 : (at your option) any later version.
13 :
14 : plumed is distributed in the hope that it will be useful,
15 : but WITHOUT ANY WARRANTY; without even the implied warranty of
16 : MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17 : GNU Lesser General Public License for more details.
18 :
19 : You should have received a copy of the GNU Lesser General Public License
20 : along with plumed. If not, see <http://www.gnu.org/licenses/>.
21 : +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ */
22 :
23 : #include "bias/Bias.h"
24 : #include "core/ActionRegister.h"
25 : #include "core/PlumedMain.h"
26 : #include "core/Value.h"
27 : #include "tools/File.h"
28 : #include "tools/OpenMP.h"
29 : #include "tools/Random.h"
30 : #include "tools/Communicator.h"
31 : #include <chrono>
32 : #include <numeric>
33 :
34 : #ifndef M_PI
35 : #define M_PI 3.14159265358979323846
36 : #endif
37 :
38 : namespace PLMD {
39 : namespace isdb {
40 :
41 : //+PLUMEDOC ISDB_BIAS METAINFERENCE
42 : /*
43 : Calculates the Metainference energy for a set of experimental data.
44 :
45 : The metainference method that is introduced in the first paper cited below is a Bayesian framework
46 : to model heterogeneous systems by integrating prior information with noisy, ensemble-averaged data.
47 : Metainference models a system and quantifies the level of noise in the data by considering a set of replicas of the system.
48 :
49 : Calculated experimental data are given in input as ARG while reference experimental values
50 : can be given either from fixed components of other actions using PARARG or as numbers using
51 : PARAMETERS. The default behavior is that of averaging the data over the available replicas,
52 : if this is not wanted the keyword NOENSEMBLE prevent this averaging.
53 :
54 : Metadynamics Metainference (see second paper cited below) more in general biased Metainference requires the knowledge of
55 : biasing potential in order to calculate the weighted average. In this case the value of the bias
56 : can be provided as the last argument in ARG and adding the keyword REWEIGHT. To avoid the noise
57 : resulting from the instantaneous value of the bias the weight of each replica can be averaged
58 : over a give time using the keyword AVERAGING.
59 :
60 : The data can be averaged by using multiple replicas and weighted for a bias if present.
61 : The functional form of Metainference can be chosen among four variants selected
62 : with NOISE=GAUSS,MGAUSS,OUTLIERS,MOUTLIERS,GENERIC which correspond to modelling the noise for
63 : the arguments as a single gaussian common to all the data points, a gaussian per data
64 : point, a single long-tailed gaussian common to all the data points, a log-tailed
65 : gaussian per data point or using two distinct noises as for the most general formulation of Metainference.
66 : In this latter case the noise of the replica-averaging is gaussian (one per data point) and the noise for
67 : the comparison with the experimental data can chosen using the keyword LIKELIHOOD
68 : between gaussian or log-normal (one per data point), furthermore the evolution of the estimated average
69 : over an infinite number of replicas is driven by DFTILDE.
70 :
71 : As for Metainference theory there are two sigma values: SIGMA_MEAN0 represent the
72 : error of calculating an average quantity using a finite set of replica and should
73 : be set as small as possible following the guidelines for replica-averaged simulations
74 : in the framework of the Maximum Entropy Principle. Alternatively, this can be obtained
75 : automatically using the internal sigma mean optimization as introduced in the third paper cited below
76 : (OPTSIGMAMEAN=SEM), in this second case sigma_mean is estimated from the maximum standard error
77 : of the mean either over the simulation or over a defined time using the keyword AVERAGING.
78 : SIGMA_BIAS is an uncertainty parameter, sampled by a MC algorithm in the bounded interval
79 : defined by SIGMA_MIN and SIGMA_MAX. The initial value is set at SIGMA0. The MC move is a
80 : random displacement of maximum value equal to DSIGMA. If the number of data point is
81 : too large and the acceptance rate drops it is possible to make the MC move over mutually
82 : exclusive, random subset of size MC_CHUNKSIZE and run more than one move setting MC_STEPS
83 : in such a way that MC_CHUNKSIZE*MC_STEPS will cover all the data points.
84 :
85 : Calculated and experimental data can be compared modulo a scaling factor and/or an offset
86 : using SCALEDATA and/or ADDOFFSET, the sampling is obtained by a MC algorithm either using
87 : a flat or a gaussian prior setting it with SCALE_PRIOR or OFFSET_PRIOR.
88 :
89 : ## Examples
90 :
91 : In the following example we calculate a set of [RDC](RDC.md), take the replica-average of
92 : them and comparing them with a set of experimental values. RDCs are compared with
93 : the experimental data but for a multiplication factor SCALE that is also sampled by
94 : MC on-the-fly
95 :
96 : ```plumed
97 : RDC ...
98 : LABEL=rdc
99 : SCALE=0.0001
100 : GYROM=-72.5388
101 : ATOMS1=22,23
102 : ATOMS2=25,27
103 : ATOMS3=29,31
104 : ATOMS4=33,34
105 : ... RDC
106 :
107 : METAINFERENCE ...
108 : ARG=rdc.*
109 : NOISETYPE=MGAUSS
110 : PARAMETERS=1.9190,2.9190,3.9190,4.9190
111 : SCALEDATA SCALE0=1 SCALE_MIN=0.1 SCALE_MAX=3 DSCALE=0.01
112 : SIGMA0=0.01 SIGMA_MIN=0.00001 SIGMA_MAX=3 DSIGMA=0.01
113 : SIGMA_MEAN0=0.001
114 : LABEL=spe
115 : ... METAINFERENCE
116 :
117 : PRINT ARG=spe.bias FILE=BIAS STRIDE=1
118 : ```
119 :
120 : in the following example instead of using one uncertainty parameter per data point we use
121 : a single uncertainty value in a long-tailed gaussian to take into account for outliers, furthermore
122 : the data are weighted for the bias applied to other variables of the system.
123 :
124 : ```plumed
125 : RDC ...
126 : LABEL=rdc
127 : SCALE=0.0001
128 : GYROM=-72.5388
129 : ATOMS1=22,23
130 : ATOMS2=25,27
131 : ATOMS3=29,31
132 : ATOMS4=33,34
133 : ... RDC
134 :
135 : cv1: TORSION ATOMS=1,2,3,4
136 : cv2: TORSION ATOMS=2,3,4,5
137 : mm: METAD ARG=cv1,cv2 HEIGHT=0.5 SIGMA=0.3,0.3 PACE=200 BIASFACTOR=8 WALKERS_MPI
138 :
139 : METAINFERENCE ...
140 : #SETTINGS NREPLICAS=2
141 : ARG=rdc.*,mm.bias
142 : REWEIGHT
143 : NOISETYPE=OUTLIERS
144 : PARAMETERS=1.9190,2.9190,3.9190,4.9190
145 : SCALEDATA SCALE0=1 SCALE_MIN=0.1 SCALE_MAX=3 DSCALE=0.01
146 : SIGMA0=0.01 SIGMA_MIN=0.00001 SIGMA_MAX=3 DSIGMA=0.01
147 : SIGMA_MEAN0=0.001
148 : LABEL=spe
149 : ... METAINFERENCE
150 : ```
151 :
152 : (See also [RDC](RDC.md), [PBMETAD](PBMETAD.md)).
153 :
154 : */
155 : //+ENDPLUMEDOC
156 :
157 : class Metainference : public bias::Bias {
158 : // experimental values
159 : std::vector<double> parameters;
160 : // noise type
161 : unsigned noise_type_;
162 : enum { GAUSS, MGAUSS, OUTLIERS, MOUTLIERS, GENERIC };
163 : unsigned gen_likelihood_;
164 : enum { LIKE_GAUSS, LIKE_LOGN };
165 : // scale is data scaling factor
166 : // noise type
167 : unsigned scale_prior_;
168 : enum { SC_GAUSS, SC_FLAT };
169 : bool doscale_;
170 : double scale_;
171 : double scale_mu_;
172 : double scale_min_;
173 : double scale_max_;
174 : double Dscale_;
175 : // scale is data scaling factor
176 : // noise type
177 : unsigned offset_prior_;
178 : bool dooffset_;
179 : double offset_;
180 : double offset_mu_;
181 : double offset_min_;
182 : double offset_max_;
183 : double Doffset_;
184 : // scale and offset regression
185 : bool doregres_zero_;
186 : int nregres_zero_;
187 : // sigma is data uncertainty
188 : std::vector<double> sigma_;
189 : std::vector<double> sigma_min_;
190 : std::vector<double> sigma_max_;
191 : std::vector<double> Dsigma_;
192 : // sigma_mean is uncertainty in the mean estimate
193 : std::vector<double> sigma_mean2_;
194 : // this is the estimator of the mean value per replica for generic metainference
195 : std::vector<double> ftilde_;
196 : double Dftilde_;
197 :
198 : // temperature in kbt
199 : double kbt_;
200 :
201 : // Monte Carlo stuff
202 : std::vector<Random> random;
203 : unsigned MCsteps_;
204 : long long unsigned MCaccept_;
205 : long long unsigned MCacceptScale_;
206 : long long unsigned MCacceptFT_;
207 : long long unsigned MCtrial_;
208 : unsigned MCchunksize_;
209 :
210 : // output
211 : Value* valueScale;
212 : Value* valueOffset;
213 : Value* valueAccept;
214 : Value* valueAcceptScale;
215 : Value* valueAcceptFT;
216 : std::vector<Value*> valueSigma;
217 : std::vector<Value*> valueSigmaMean;
218 : std::vector<Value*> valueFtilde;
219 :
220 : // restart
221 : unsigned write_stride_;
222 : OFile sfile_;
223 :
224 : // others
225 : bool firstTime;
226 : std::vector<bool> firstTimeW;
227 : bool master;
228 : bool do_reweight_;
229 : unsigned do_optsigmamean_;
230 : unsigned nrep_;
231 : unsigned replica_;
232 : unsigned narg;
233 :
234 : // selector
235 : std::string selector_;
236 :
237 : // optimize sigma mean
238 : std::vector< std::vector < std::vector <double> > > sigma_mean2_last_;
239 : unsigned optsigmamean_stride_;
240 : // optimize sigma max
241 : unsigned N_optimized_step_;
242 : unsigned optimized_step_;
243 : bool sigmamax_opt_done_;
244 : std::vector<double> sigma_max_est_;
245 :
246 : // average weights
247 : unsigned average_weights_stride_;
248 : std::vector< std::vector <double> > average_weights_;
249 :
250 : double getEnergyMIGEN(const std::vector<double> &mean, const std::vector<double> &ftilde, const std::vector<double> &sigma,
251 : const double scale, const double offset);
252 : double getEnergySP(const std::vector<double> &mean, const std::vector<double> &sigma,
253 : const double scale, const double offset);
254 : double getEnergySPE(const std::vector<double> &mean, const std::vector<double> &sigma,
255 : const double scale, const double offset);
256 : double getEnergyGJ(const std::vector<double> &mean, const std::vector<double> &sigma,
257 : const double scale, const double offset);
258 : double getEnergyGJE(const std::vector<double> &mean, const std::vector<double> &sigma,
259 : const double scale, const double offset);
260 : void moveTilde(const std::vector<double> &mean_, double &old_energy);
261 : void moveScaleOffset(const std::vector<double> &mean_, double &old_energy);
262 : void moveSigmas(const std::vector<double> &mean_, double &old_energy, const unsigned i, const std::vector<unsigned> &indices, bool &breaknow);
263 : double doMonteCarlo(const std::vector<double> &mean);
264 : void getEnergyForceMIGEN(const std::vector<double> &mean, const std::vector<double> &dmean_x, const std::vector<double> &dmean_b);
265 : void getEnergyForceSP(const std::vector<double> &mean, const std::vector<double> &dmean_x, const std::vector<double> &dmean_b);
266 : void getEnergyForceSPE(const std::vector<double> &mean, const std::vector<double> &dmean_x, const std::vector<double> &dmean_b);
267 : void getEnergyForceGJ(const std::vector<double> &mean, const std::vector<double> &dmean_x, const std::vector<double> &dmean_b);
268 : void getEnergyForceGJE(const std::vector<double> &mean, const std::vector<double> &dmean_x, const std::vector<double> &dmean_b);
269 : void get_weights(const unsigned iselect, double &weight, double &norm, double &neff);
270 : void replica_averaging(const double weight, const double norm, std::vector<double> &mean, std::vector<double> &dmean_b);
271 : void get_sigma_mean(const unsigned iselect, const double weight, const double norm, const double neff, const std::vector<double> &mean);
272 : void writeStatus();
273 : void do_regression_zero(const std::vector<double> &mean);
274 :
275 : public:
276 : explicit Metainference(const ActionOptions&);
277 : ~Metainference();
278 : void calculate() override;
279 : void update() override;
280 : static void registerKeywords(Keywords& keys);
281 : };
282 :
283 :
284 : PLUMED_REGISTER_ACTION(Metainference,"METAINFERENCE")
285 :
286 34 : void Metainference::registerKeywords(Keywords& keys) {
287 34 : Bias::registerKeywords(keys);
288 68 : keys.addInputKeyword("optional","PARARG","scalar","reference values for the experimental data, these can be provided as arguments without derivatives");
289 34 : keys.add("optional","PARAMETERS","reference values for the experimental data");
290 34 : keys.addFlag("NOENSEMBLE",false,"don't perform any replica-averaging");
291 34 : keys.addFlag("REWEIGHT",false,"simple REWEIGHT using the latest ARG as energy");
292 34 : keys.add("optional","AVERAGING", "Stride for calculation of averaged weights and sigma_mean");
293 34 : keys.add("compulsory","NOISETYPE","MGAUSS","functional form of the noise (GAUSS,MGAUSS,OUTLIERS,MOUTLIERS,GENERIC)");
294 34 : keys.add("compulsory","LIKELIHOOD","GAUSS","the likelihood for the GENERIC metainference model, GAUSS or LOGN");
295 34 : keys.add("compulsory","DFTILDE","0.1","fraction of sigma_mean used to evolve ftilde");
296 34 : keys.addFlag("SCALEDATA",false,"Set to TRUE if you want to sample a scaling factor common to all values and replicas");
297 34 : keys.add("compulsory","SCALE0","1.0","initial value of the scaling factor");
298 34 : keys.add("compulsory","SCALE_PRIOR","FLAT","either FLAT or GAUSSIAN");
299 34 : keys.add("optional","SCALE_MIN","minimum value of the scaling factor");
300 34 : keys.add("optional","SCALE_MAX","maximum value of the scaling factor");
301 34 : keys.add("optional","DSCALE","maximum MC move of the scaling factor");
302 34 : keys.addFlag("ADDOFFSET",false,"Set to TRUE if you want to sample an offset common to all values and replicas");
303 34 : keys.add("compulsory","OFFSET0","0.0","initial value of the offset");
304 34 : keys.add("compulsory","OFFSET_PRIOR","FLAT","either FLAT or GAUSSIAN");
305 34 : keys.add("optional","OFFSET_MIN","minimum value of the offset");
306 34 : keys.add("optional","OFFSET_MAX","maximum value of the offset");
307 34 : keys.add("optional","DOFFSET","maximum MC move of the offset");
308 34 : keys.add("optional","REGRES_ZERO","stride for regression with zero offset");
309 34 : keys.add("compulsory","SIGMA0","1.0","initial value of the uncertainty parameter");
310 34 : keys.add("compulsory","SIGMA_MIN","0.0","minimum value of the uncertainty parameter");
311 34 : keys.add("compulsory","SIGMA_MAX","10.","maximum value of the uncertainty parameter");
312 34 : keys.add("optional","DSIGMA","maximum MC move of the uncertainty parameter");
313 34 : keys.add("compulsory","OPTSIGMAMEAN","NONE","Set to NONE/SEM to manually set sigma mean, or to estimate it on the fly");
314 34 : keys.add("optional","SIGMA_MEAN0","starting value for the uncertainty in the mean estimate");
315 34 : keys.add("optional","SIGMA_MAX_STEPS", "Number of steps used to optimise SIGMA_MAX, before that the SIGMA_MAX value is used");
316 34 : keys.add("optional","TEMP","the system temperature - this is only needed if code doesn't pass the temperature to plumed");
317 34 : keys.add("optional","MC_STEPS","number of MC steps");
318 34 : keys.add("optional","MC_CHUNKSIZE","MC chunksize");
319 34 : keys.add("optional","STATUS_FILE","write a file with all the data useful for restart/continuation of Metainference");
320 34 : keys.add("optional","FMT","specify format for HILLS files (useful for decrease the number of digits in regtests)");
321 34 : keys.add("compulsory","WRITE_STRIDE","10000","write the status to a file every N steps, this can be used for restart/continuation");
322 34 : keys.add("optional","SELECTOR","name of selector");
323 34 : keys.add("optional","NSELECT","range of values for selector [0, N-1]");
324 34 : keys.use("RESTART");
325 68 : keys.addOutputComponent("sigma", "default", "scalar","uncertainty parameter");
326 68 : keys.addOutputComponent("sigmaMean", "default", "scalar","uncertainty in the mean estimate");
327 68 : keys.addOutputComponent("neff", "default", "scalar","effective number of replicas");
328 68 : keys.addOutputComponent("acceptSigma", "default", "scalar","MC acceptance for sigma values");
329 68 : keys.addOutputComponent("acceptScale", "SCALEDATA", "scalar","MC acceptance for scale value");
330 68 : keys.addOutputComponent("acceptFT", "GENERIC", "scalar","MC acceptance for general metainference f tilde value");
331 68 : keys.addOutputComponent("weight", "REWEIGHT", "scalar","weights of the weighted average");
332 68 : keys.addOutputComponent("biasDer", "REWEIGHT", "scalar","derivatives with respect to the bias");
333 68 : keys.addOutputComponent("scale", "SCALEDATA", "scalar","scale parameter");
334 68 : keys.addOutputComponent("offset", "ADDOFFSET", "scalar","offset parameter");
335 68 : keys.addOutputComponent("ftilde", "GENERIC", "scalar","ensemble average estimator");
336 34 : keys.addDOI("10.1126/sciadv.1501177");
337 34 : keys.addDOI("10.1038/srep31232");
338 34 : keys.addDOI("10.1063/1.4981211");
339 34 : }
340 :
341 32 : Metainference::Metainference(const ActionOptions&ao):
342 : PLUMED_BIAS_INIT(ao),
343 32 : doscale_(false),
344 32 : scale_(1.),
345 32 : scale_mu_(0),
346 32 : scale_min_(1),
347 32 : scale_max_(-1),
348 32 : Dscale_(-1),
349 32 : dooffset_(false),
350 32 : offset_(0.),
351 32 : offset_mu_(0),
352 32 : offset_min_(1),
353 32 : offset_max_(-1),
354 32 : Doffset_(-1),
355 32 : doregres_zero_(false),
356 32 : nregres_zero_(0),
357 32 : Dftilde_(0.1),
358 32 : random(3),
359 32 : MCsteps_(1),
360 32 : MCaccept_(0),
361 32 : MCacceptScale_(0),
362 32 : MCacceptFT_(0),
363 32 : MCtrial_(0),
364 32 : MCchunksize_(0),
365 32 : write_stride_(0),
366 32 : firstTime(true),
367 32 : do_reweight_(false),
368 32 : do_optsigmamean_(0),
369 32 : optsigmamean_stride_(0),
370 32 : N_optimized_step_(0),
371 32 : optimized_step_(0),
372 32 : sigmamax_opt_done_(false),
373 32 : average_weights_stride_(1) {
374 32 : bool noensemble = false;
375 32 : parseFlag("NOENSEMBLE", noensemble);
376 :
377 : // set up replica stuff
378 32 : master = (comm.Get_rank()==0);
379 32 : if(master) {
380 24 : nrep_ = multi_sim_comm.Get_size();
381 24 : replica_ = multi_sim_comm.Get_rank();
382 24 : if(noensemble) {
383 0 : nrep_ = 1;
384 : }
385 : } else {
386 8 : nrep_ = 0;
387 8 : replica_ = 0;
388 : }
389 32 : comm.Sum(&nrep_,1);
390 32 : comm.Sum(&replica_,1);
391 :
392 32 : unsigned nsel = 1;
393 32 : parse("SELECTOR", selector_);
394 64 : parse("NSELECT", nsel);
395 : // do checks
396 32 : if(selector_.length()>0 && nsel<=1) {
397 0 : error("With SELECTOR active, NSELECT must be greater than 1");
398 : }
399 32 : if(selector_.length()==0 && nsel>1) {
400 0 : error("With NSELECT greater than 1, you must specify SELECTOR");
401 : }
402 :
403 : // initialise firstTimeW
404 32 : firstTimeW.resize(nsel, true);
405 :
406 : // reweight implies a different number of arguments (the latest one must always be the bias)
407 32 : parseFlag("REWEIGHT", do_reweight_);
408 32 : if(do_reweight_&&nrep_<2) {
409 0 : error("REWEIGHT can only be used in parallel with 2 or more replicas");
410 : }
411 32 : if(!getRestart()) {
412 56 : average_weights_.resize(nsel, std::vector<double> (nrep_, 1./static_cast<double>(nrep_)));
413 : } else {
414 8 : average_weights_.resize(nsel, std::vector<double> (nrep_, 0.));
415 : }
416 32 : narg = getNumberOfArguments();
417 32 : if(do_reweight_) {
418 16 : narg--;
419 : }
420 :
421 32 : unsigned averaging=0;
422 32 : parse("AVERAGING", averaging);
423 32 : if(averaging>0) {
424 1 : average_weights_stride_ = averaging;
425 1 : optsigmamean_stride_ = averaging;
426 : }
427 :
428 64 : parseVector("PARAMETERS",parameters);
429 32 : if(parameters.size()!=static_cast<unsigned>(narg)&&!parameters.empty()) {
430 0 : error("Size of PARAMETERS array should be either 0 or the same as of the number of arguments in ARG1");
431 : }
432 :
433 : std::vector<Value*> arg2;
434 64 : parseArgumentList("PARARG",arg2);
435 32 : if(!arg2.empty()) {
436 4 : if(parameters.size()>0) {
437 0 : error("It is not possible to use PARARG and PARAMETERS together");
438 : }
439 4 : if(arg2.size()!=narg) {
440 0 : error("Size of PARARG array should be the same as number for arguments in ARG");
441 : }
442 2360 : for(unsigned i=0; i<arg2.size(); i++) {
443 2356 : parameters.push_back(arg2[i]->get());
444 2356 : if(arg2[i]->hasDerivatives()==true) {
445 0 : error("PARARG can only accept arguments without derivatives");
446 : }
447 : }
448 : }
449 :
450 32 : if(parameters.size()!=narg) {
451 0 : error("PARARG or PARAMETERS arrays should include the same number of elements as the arguments in ARG");
452 : }
453 : {
454 : std::string stringa_noise;
455 64 : parse("NOISETYPE",stringa_noise);
456 32 : if(stringa_noise=="GAUSS") {
457 3 : noise_type_ = GAUSS;
458 29 : } else if(stringa_noise=="MGAUSS") {
459 8 : noise_type_ = MGAUSS;
460 21 : } else if(stringa_noise=="OUTLIERS") {
461 14 : noise_type_ = OUTLIERS;
462 7 : } else if(stringa_noise=="MOUTLIERS") {
463 4 : noise_type_ = MOUTLIERS;
464 3 : } else if(stringa_noise=="GENERIC") {
465 3 : noise_type_ = GENERIC;
466 : } else {
467 0 : error("Unknown noise type!");
468 : }
469 : }
470 32 : if(noise_type_== GENERIC) {
471 : std::string stringa_like;
472 6 : parse("LIKELIHOOD",stringa_like);
473 3 : if(stringa_like=="GAUSS") {
474 2 : gen_likelihood_ = LIKE_GAUSS;
475 1 : } else if(stringa_like=="LOGN") {
476 1 : gen_likelihood_ = LIKE_LOGN;
477 : } else {
478 0 : error("Unknown likelihood type!");
479 : }
480 :
481 6 : parse("DFTILDE",Dftilde_);
482 : }
483 :
484 64 : parse("WRITE_STRIDE",write_stride_);
485 : std::string status_file_name_;
486 64 : parse("STATUS_FILE",status_file_name_);
487 32 : if(status_file_name_=="") {
488 64 : status_file_name_ = "MISTATUS"+getLabel();
489 : } else {
490 0 : status_file_name_ = status_file_name_+getLabel();
491 : }
492 :
493 : std::string stringa_optsigma;
494 64 : parse("OPTSIGMAMEAN", stringa_optsigma);
495 32 : if(stringa_optsigma=="NONE") {
496 30 : do_optsigmamean_=0;
497 2 : } else if(stringa_optsigma=="SEM") {
498 0 : do_optsigmamean_=1;
499 2 : } else if(stringa_optsigma=="SEM_MAX") {
500 0 : do_optsigmamean_=2;
501 : }
502 :
503 32 : unsigned aver_max_steps=0;
504 32 : parse("SIGMA_MAX_STEPS", aver_max_steps);
505 32 : if(aver_max_steps==0&&do_optsigmamean_==2) {
506 0 : aver_max_steps=averaging*2000;
507 : }
508 32 : if(aver_max_steps>0&&do_optsigmamean_<2) {
509 0 : error("SIGMA_MAX_STEPS can only be used together with OPTSIGMAMEAN=SEM_MAX");
510 : }
511 32 : if(aver_max_steps>0&&do_optsigmamean_==2) {
512 0 : N_optimized_step_=aver_max_steps;
513 : }
514 32 : if(aver_max_steps>0&&aver_max_steps<averaging) {
515 0 : error("SIGMA_MAX_STEPS must be greater than AVERAGING");
516 : }
517 :
518 : // resize std::vector for sigma_mean history
519 32 : sigma_mean2_last_.resize(nsel);
520 64 : for(unsigned i=0; i<nsel; i++) {
521 32 : sigma_mean2_last_[i].resize(narg);
522 : }
523 :
524 : std::vector<double> read_sigma_mean_;
525 32 : parseVector("SIGMA_MEAN0",read_sigma_mean_);
526 32 : if(do_optsigmamean_==0 && read_sigma_mean_.size()==0 && !getRestart()) {
527 0 : error("If you don't use OPTSIGMAMEAN and you are not RESTARTING then you MUST SET SIGMA_MEAN0");
528 : }
529 :
530 32 : if(noise_type_==MGAUSS||noise_type_==MOUTLIERS||noise_type_==GENERIC) {
531 15 : if(read_sigma_mean_.size()==narg) {
532 0 : sigma_mean2_.resize(narg);
533 0 : for(unsigned i=0; i<narg; i++) {
534 0 : sigma_mean2_[i]=read_sigma_mean_[i]*read_sigma_mean_[i];
535 : }
536 15 : } else if(read_sigma_mean_.size()==1) {
537 15 : sigma_mean2_.resize(narg,read_sigma_mean_[0]*read_sigma_mean_[0]);
538 0 : } else if(read_sigma_mean_.size()==0) {
539 0 : sigma_mean2_.resize(narg,0.000001);
540 : } else {
541 0 : error("SIGMA_MEAN0 can accept either one single value or as many values as the arguments (with NOISETYPE=MGAUSS|MOUTLIERS)");
542 : }
543 : // set the initial value for the history
544 30 : for(unsigned i=0; i<nsel; i++)
545 2409 : for(unsigned j=0; j<narg; j++) {
546 2394 : sigma_mean2_last_[i][j].push_back(sigma_mean2_[j]);
547 : }
548 : } else {
549 17 : if(read_sigma_mean_.size()==1) {
550 17 : sigma_mean2_.resize(1, read_sigma_mean_[0]*read_sigma_mean_[0]);
551 0 : } else if(read_sigma_mean_.size()==0) {
552 0 : sigma_mean2_.resize(1, 0.000001);
553 : } else {
554 0 : error("If you want to use more than one SIGMA_MEAN0 you should use NOISETYPE=MGAUSS|MOUTLIERS");
555 : }
556 : // set the initial value for the history
557 34 : for(unsigned i=0; i<nsel; i++)
558 109 : for(unsigned j=0; j<narg; j++) {
559 92 : sigma_mean2_last_[i][j].push_back(sigma_mean2_[0]);
560 : }
561 : }
562 :
563 32 : parseFlag("SCALEDATA", doscale_);
564 32 : if(doscale_) {
565 : std::string stringa_noise;
566 24 : parse("SCALE_PRIOR",stringa_noise);
567 12 : if(stringa_noise=="GAUSSIAN") {
568 0 : scale_prior_ = SC_GAUSS;
569 12 : } else if(stringa_noise=="FLAT") {
570 12 : scale_prior_ = SC_FLAT;
571 : } else {
572 0 : error("Unknown SCALE_PRIOR type!");
573 : }
574 12 : parse("SCALE0",scale_);
575 12 : parse("DSCALE",Dscale_);
576 12 : if(Dscale_<0.) {
577 0 : error("DSCALE must be set when using SCALEDATA");
578 : }
579 12 : if(scale_prior_==SC_GAUSS) {
580 0 : scale_mu_=scale_;
581 : } else {
582 12 : parse("SCALE_MIN",scale_min_);
583 12 : parse("SCALE_MAX",scale_max_);
584 12 : if(scale_max_<scale_min_) {
585 0 : error("SCALE_MAX and SCALE_MIN must be set when using SCALE_PRIOR=FLAT");
586 : }
587 : }
588 : }
589 :
590 32 : parseFlag("ADDOFFSET", dooffset_);
591 32 : if(dooffset_) {
592 : std::string stringa_noise;
593 12 : parse("OFFSET_PRIOR",stringa_noise);
594 6 : if(stringa_noise=="GAUSSIAN") {
595 0 : offset_prior_ = SC_GAUSS;
596 6 : } else if(stringa_noise=="FLAT") {
597 6 : offset_prior_ = SC_FLAT;
598 : } else {
599 0 : error("Unknown OFFSET_PRIOR type!");
600 : }
601 6 : parse("OFFSET0",offset_);
602 6 : parse("DOFFSET",Doffset_);
603 6 : if(offset_prior_==SC_GAUSS) {
604 0 : offset_mu_=offset_;
605 0 : if(Doffset_<0.) {
606 0 : error("DOFFSET must be set when using OFFSET_PRIOR=GAUSS");
607 : }
608 : } else {
609 6 : parse("OFFSET_MIN",offset_min_);
610 6 : parse("OFFSET_MAX",offset_max_);
611 6 : if(Doffset_<0) {
612 0 : Doffset_ = 0.05*(offset_max_ - offset_min_);
613 : }
614 6 : if(offset_max_<offset_min_) {
615 0 : error("OFFSET_MAX and OFFSET_MIN must be set when using OFFSET_PRIOR=FLAT");
616 : }
617 : }
618 : }
619 :
620 : // regression with zero intercept
621 32 : parse("REGRES_ZERO", nregres_zero_);
622 32 : if(nregres_zero_>0) {
623 : // set flag
624 1 : doregres_zero_=true;
625 : // check if already sampling scale and offset
626 1 : if(doscale_) {
627 0 : error("REGRES_ZERO and SCALEDATA are mutually exclusive");
628 : }
629 1 : if(dooffset_) {
630 0 : error("REGRES_ZERO and ADDOFFSET are mutually exclusive");
631 : }
632 : }
633 :
634 : std::vector<double> readsigma;
635 32 : parseVector("SIGMA0",readsigma);
636 32 : if((noise_type_!=MGAUSS&&noise_type_!=MOUTLIERS&&noise_type_!=GENERIC)&&readsigma.size()>1) {
637 0 : error("If you want to use more than one SIGMA you should use NOISETYPE=MGAUSS|MOUTLIERS|GENERIC");
638 : }
639 32 : if(noise_type_==MGAUSS||noise_type_==MOUTLIERS||noise_type_==GENERIC) {
640 15 : sigma_.resize(readsigma.size());
641 15 : sigma_=readsigma;
642 : } else {
643 17 : sigma_.resize(1, readsigma[0]);
644 : }
645 :
646 : std::vector<double> readsigma_min;
647 32 : parseVector("SIGMA_MIN",readsigma_min);
648 32 : if((noise_type_!=MGAUSS&&noise_type_!=MOUTLIERS&&noise_type_!=GENERIC)&&readsigma_min.size()>1) {
649 0 : error("If you want to use more than one SIGMA you should use NOISETYPE=MGAUSS|MOUTLIERS|GENERIC");
650 : }
651 32 : if(noise_type_==MGAUSS||noise_type_==MOUTLIERS||noise_type_==GENERIC) {
652 15 : sigma_min_.resize(readsigma_min.size());
653 15 : sigma_min_=readsigma_min;
654 : } else {
655 17 : sigma_min_.resize(1, readsigma_min[0]);
656 : }
657 :
658 : std::vector<double> readsigma_max;
659 32 : parseVector("SIGMA_MAX",readsigma_max);
660 32 : if((noise_type_!=MGAUSS&&noise_type_!=MOUTLIERS&&noise_type_!=GENERIC)&&readsigma_max.size()>1) {
661 0 : error("If you want to use more than one SIGMA you should use NOISETYPE=MGAUSS|MOUTLIERS|GENERIC");
662 : }
663 32 : if(noise_type_==MGAUSS||noise_type_==MOUTLIERS||noise_type_==GENERIC) {
664 15 : sigma_max_.resize(readsigma_max.size());
665 15 : sigma_max_=readsigma_max;
666 : } else {
667 17 : sigma_max_.resize(1, readsigma_max[0]);
668 : }
669 :
670 32 : if(sigma_max_.size()!=sigma_min_.size()) {
671 0 : error("The number of values for SIGMA_MIN and SIGMA_MAX must be the same");
672 : }
673 :
674 : std::vector<double> read_dsigma;
675 32 : parseVector("DSIGMA",read_dsigma);
676 32 : if((noise_type_!=MGAUSS&&noise_type_!=MOUTLIERS&&noise_type_!=GENERIC)&&readsigma_max.size()>1) {
677 0 : error("If you want to use more than one SIGMA you should use NOISETYPE=MGAUSS|MOUTLIERS|GENERIC");
678 : }
679 32 : if(read_dsigma.size()>0) {
680 32 : Dsigma_.resize(read_dsigma.size());
681 32 : Dsigma_=read_dsigma;
682 : } else {
683 0 : Dsigma_.resize(sigma_max_.size(), -1.);
684 : /* in this case Dsigma is initialised after reading the restart file if present */
685 : }
686 :
687 : std::string fmt_;
688 32 : parse("FMT",fmt_);
689 :
690 : // monte carlo stuff
691 32 : parse("MC_STEPS",MCsteps_);
692 32 : parse("MC_CHUNKSIZE", MCchunksize_);
693 : // get temperature
694 32 : kbt_ = getkBT();
695 32 : if(kbt_==0.0) {
696 0 : error("Unless the MD engine passes the temperature to plumed, you must specify it using TEMP");
697 : }
698 :
699 32 : checkRead();
700 :
701 : // set sigma_bias
702 32 : if(noise_type_==MGAUSS||noise_type_==MOUTLIERS||noise_type_==GENERIC) {
703 15 : if(sigma_.size()==1) {
704 15 : double tmp = sigma_[0];
705 15 : sigma_.resize(narg, tmp);
706 0 : } else if(sigma_.size()>1&&sigma_.size()!=narg) {
707 0 : error("SIGMA0 can accept either one single value or as many values as the number of arguments (with NOISETYPE=MGAUSS|MOUTLIERS|GENERIC)");
708 : }
709 15 : if(sigma_min_.size()==1) {
710 15 : double tmp = sigma_min_[0];
711 15 : sigma_min_.resize(narg, tmp);
712 0 : } else if(sigma_min_.size()>1&&sigma_min_.size()!=narg) {
713 0 : error("SIGMA_MIN can accept either one single value or as many values as the number of arguments (with NOISETYPE=MGAUSS|MOUTLIERS|GENERIC)");
714 : }
715 15 : if(sigma_max_.size()==1) {
716 15 : double tmp = sigma_max_[0];
717 15 : sigma_max_.resize(narg, tmp);
718 0 : } else if(sigma_max_.size()>1&&sigma_max_.size()!=narg) {
719 0 : error("SIGMA_MAX can accept either one single value or as many values as the number of arguments (with NOISETYPE=MGAUSS|MOUTLIERS|GENERIC)");
720 : }
721 15 : if(Dsigma_.size()==1) {
722 15 : double tmp = Dsigma_[0];
723 15 : Dsigma_.resize(narg, tmp);
724 0 : } else if(Dsigma_.size()>1&&Dsigma_.size()!=narg) {
725 0 : error("DSIGMA can accept either one single value or as many values as the number of arguments (with NOISETYPE=MGAUSS|MOUTLIERS|GENERIC)");
726 : }
727 : }
728 :
729 32 : sigma_max_est_.resize(sigma_max_.size(), 0.);
730 :
731 32 : IFile restart_sfile;
732 32 : restart_sfile.link(*this);
733 32 : if(getRestart()&&restart_sfile.FileExist(status_file_name_)) {
734 4 : firstTime = false;
735 8 : for(unsigned i=0; i<nsel; i++) {
736 : firstTimeW[i] = false;
737 : }
738 4 : restart_sfile.open(status_file_name_);
739 4 : log.printf(" Restarting from %s\n", status_file_name_.c_str());
740 : double dummy;
741 8 : if(restart_sfile.scanField("time",dummy)) {
742 : // check for syncronisation
743 4 : std::vector<double> dummy_time(nrep_,0);
744 4 : if(master&&nrep_>1) {
745 2 : dummy_time[replica_] = dummy;
746 2 : multi_sim_comm.Sum(dummy_time);
747 : }
748 4 : comm.Sum(dummy_time);
749 8 : for(unsigned i=1; i<nrep_; i++) {
750 4 : std::string msg = "METAINFERENCE restart files " + status_file_name_ + " are not in sync";
751 4 : if(dummy_time[i]!=dummy_time[0]) {
752 0 : plumed_merror(msg);
753 : }
754 : }
755 : // nsel
756 8 : for(unsigned i=0; i<sigma_mean2_last_.size(); i++) {
757 : std::string msg_i;
758 4 : Tools::convert(i,msg_i);
759 : // narg
760 4 : if(noise_type_==MGAUSS||noise_type_==MOUTLIERS||noise_type_==GENERIC) {
761 20 : for(unsigned j=0; j<narg; ++j) {
762 : std::string msg_j;
763 16 : Tools::convert(j,msg_j);
764 16 : std::string msg = msg_i+"_"+msg_j;
765 : double read_sm;
766 16 : restart_sfile.scanField("sigmaMean_"+msg,read_sm);
767 16 : sigma_mean2_last_[i][j][0] = read_sm*read_sm;
768 : }
769 : }
770 4 : if(noise_type_==GAUSS||noise_type_==OUTLIERS) {
771 : double read_sm;
772 : std::string msg_j;
773 0 : Tools::convert(0,msg_j);
774 0 : std::string msg = msg_i+"_"+msg_j;
775 0 : restart_sfile.scanField("sigmaMean_"+msg,read_sm);
776 0 : for(unsigned j=0; j<narg; j++) {
777 0 : sigma_mean2_last_[i][j][0] = read_sm*read_sm;
778 : }
779 : }
780 : }
781 :
782 20 : for(unsigned i=0; i<sigma_.size(); ++i) {
783 : std::string msg;
784 16 : Tools::convert(i,msg);
785 32 : restart_sfile.scanField("sigma_"+msg,sigma_[i]);
786 : }
787 20 : for(unsigned i=0; i<sigma_max_.size(); ++i) {
788 : std::string msg;
789 16 : Tools::convert(i,msg);
790 16 : restart_sfile.scanField("sigma_max_"+msg,sigma_max_[i]);
791 16 : sigmamax_opt_done_=true;
792 : }
793 4 : if(noise_type_==GENERIC) {
794 0 : for(unsigned i=0; i<ftilde_.size(); ++i) {
795 : std::string msg;
796 0 : Tools::convert(i,msg);
797 0 : restart_sfile.scanField("ftilde_"+msg,ftilde_[i]);
798 : }
799 : }
800 4 : restart_sfile.scanField("scale0_",scale_);
801 4 : restart_sfile.scanField("offset0_",offset_);
802 :
803 8 : for(unsigned i=0; i<nsel; i++) {
804 : std::string msg;
805 4 : Tools::convert(i,msg);
806 : double tmp_w;
807 4 : restart_sfile.scanField("weight_"+msg,tmp_w);
808 4 : if(master) {
809 2 : average_weights_[i][replica_] = tmp_w;
810 2 : if(nrep_>1) {
811 2 : multi_sim_comm.Sum(&average_weights_[i][0], nrep_);
812 : }
813 : }
814 4 : comm.Sum(&average_weights_[i][0], nrep_);
815 : }
816 :
817 : }
818 4 : restart_sfile.scanField();
819 4 : restart_sfile.close();
820 : }
821 :
822 : /* If DSIGMA is not yet initialised do it now */
823 2443 : for(unsigned i=0; i<sigma_max_.size(); i++)
824 2411 : if(Dsigma_[i]==-1) {
825 0 : Dsigma_[i] = 0.05*(sigma_max_[i] - sigma_min_[i]);
826 : }
827 :
828 32 : switch(noise_type_) {
829 3 : case GENERIC:
830 3 : log.printf(" with general metainference ");
831 3 : if(gen_likelihood_==LIKE_GAUSS) {
832 2 : log.printf(" and a gaussian likelihood\n");
833 1 : } else if(gen_likelihood_==LIKE_LOGN) {
834 1 : log.printf(" and a log-normal likelihood\n");
835 : }
836 3 : log.printf(" ensemble average parameter sampled with a step %lf of sigma_mean\n", Dftilde_);
837 : break;
838 3 : case GAUSS:
839 3 : log.printf(" with gaussian noise and a single noise parameter for all the data\n");
840 : break;
841 8 : case MGAUSS:
842 8 : log.printf(" with gaussian noise and a noise parameter for each data point\n");
843 : break;
844 14 : case OUTLIERS:
845 14 : log.printf(" with long tailed gaussian noise and a single noise parameter for all the data\n");
846 : break;
847 4 : case MOUTLIERS:
848 4 : log.printf(" with long tailed gaussian noise and a noise parameter for each data point\n");
849 : break;
850 : }
851 :
852 32 : if(doscale_) {
853 : // check that the scale value is the same for all replicas
854 12 : std::vector<double> dummy_scale(nrep_,0);
855 12 : if(master&&nrep_>1) {
856 6 : dummy_scale[replica_] = scale_;
857 6 : multi_sim_comm.Sum(dummy_scale);
858 : }
859 12 : comm.Sum(dummy_scale);
860 24 : for(unsigned i=1; i<nrep_; i++) {
861 12 : std::string msg = "The SCALE value must be the same for all replicas: check your input or restart file";
862 12 : if(dummy_scale[i]!=dummy_scale[0]) {
863 0 : plumed_merror(msg);
864 : }
865 : }
866 12 : log.printf(" sampling a common scaling factor with:\n");
867 12 : log.printf(" initial scale parameter %f\n",scale_);
868 12 : if(scale_prior_==SC_GAUSS) {
869 0 : log.printf(" gaussian prior with mean %f and width %f\n",scale_mu_,Dscale_);
870 : }
871 12 : if(scale_prior_==SC_FLAT) {
872 12 : log.printf(" flat prior between %f - %f\n",scale_min_,scale_max_);
873 12 : log.printf(" maximum MC move of scale parameter %f\n",Dscale_);
874 : }
875 : }
876 :
877 32 : if(dooffset_) {
878 : // check that the offset value is the same for all replicas
879 6 : std::vector<double> dummy_offset(nrep_,0);
880 6 : if(master&&nrep_>1) {
881 0 : dummy_offset[replica_] = offset_;
882 0 : multi_sim_comm.Sum(dummy_offset);
883 : }
884 6 : comm.Sum(dummy_offset);
885 6 : for(unsigned i=1; i<nrep_; i++) {
886 0 : std::string msg = "The OFFSET value must be the same for all replicas: check your input or restart file";
887 0 : if(dummy_offset[i]!=dummy_offset[0]) {
888 0 : plumed_merror(msg);
889 : }
890 : }
891 6 : log.printf(" sampling a common offset with:\n");
892 6 : log.printf(" initial offset parameter %f\n",offset_);
893 6 : if(offset_prior_==SC_GAUSS) {
894 0 : log.printf(" gaussian prior with mean %f and width %f\n",offset_mu_,Doffset_);
895 : }
896 6 : if(offset_prior_==SC_FLAT) {
897 6 : log.printf(" flat prior between %f - %f\n",offset_min_,offset_max_);
898 6 : log.printf(" maximum MC move of offset parameter %f\n",Doffset_);
899 : }
900 : }
901 :
902 32 : if(doregres_zero_) {
903 1 : log.printf(" doing regression with zero intercept with stride: %d\n", nregres_zero_);
904 : }
905 :
906 32 : log.printf(" number of experimental data points %u\n",narg);
907 32 : log.printf(" number of replicas %u\n",nrep_);
908 32 : log.printf(" initial data uncertainties");
909 2443 : for(unsigned i=0; i<sigma_.size(); ++i) {
910 2411 : log.printf(" %f", sigma_[i]);
911 : }
912 32 : log.printf("\n");
913 32 : log.printf(" minimum data uncertainties");
914 2443 : for(unsigned i=0; i<sigma_.size(); ++i) {
915 2411 : log.printf(" %f",sigma_min_[i]);
916 : }
917 32 : log.printf("\n");
918 32 : log.printf(" maximum data uncertainties");
919 2475 : for(unsigned i=0; i<sigma_.size(); ++i) {
920 2411 : log.printf(" %f",sigma_max_[i]);
921 : }
922 32 : log.printf("\n");
923 32 : log.printf(" maximum MC move of data uncertainties");
924 2443 : for(unsigned i=0; i<sigma_.size(); ++i) {
925 2411 : log.printf(" %f",Dsigma_[i]);
926 : }
927 32 : log.printf("\n");
928 32 : log.printf(" temperature of the system %f\n",kbt_);
929 32 : log.printf(" MC steps %u\n",MCsteps_);
930 32 : log.printf(" initial standard errors of the mean");
931 2443 : for(unsigned i=0; i<sigma_mean2_.size(); ++i) {
932 2411 : log.printf(" %f", std::sqrt(sigma_mean2_[i]));
933 : }
934 32 : log.printf("\n");
935 :
936 32 : if(do_reweight_) {
937 32 : addComponent("biasDer");
938 32 : componentIsNotPeriodic("biasDer");
939 32 : addComponent("weight");
940 32 : componentIsNotPeriodic("weight");
941 : }
942 :
943 64 : addComponent("neff");
944 32 : componentIsNotPeriodic("neff");
945 :
946 32 : if(doscale_ || doregres_zero_) {
947 26 : addComponent("scale");
948 13 : componentIsNotPeriodic("scale");
949 13 : valueScale=getPntrToComponent("scale");
950 : }
951 :
952 32 : if(dooffset_) {
953 12 : addComponent("offset");
954 6 : componentIsNotPeriodic("offset");
955 6 : valueOffset=getPntrToComponent("offset");
956 : }
957 :
958 32 : if(dooffset_||doscale_) {
959 36 : addComponent("acceptScale");
960 18 : componentIsNotPeriodic("acceptScale");
961 18 : valueAcceptScale=getPntrToComponent("acceptScale");
962 : }
963 :
964 32 : if(noise_type_==GENERIC) {
965 6 : addComponent("acceptFT");
966 3 : componentIsNotPeriodic("acceptFT");
967 3 : valueAcceptFT=getPntrToComponent("acceptFT");
968 : }
969 :
970 64 : addComponent("acceptSigma");
971 32 : componentIsNotPeriodic("acceptSigma");
972 32 : valueAccept=getPntrToComponent("acceptSigma");
973 :
974 32 : if(noise_type_==MGAUSS||noise_type_==MOUTLIERS||noise_type_==GENERIC) {
975 2409 : for(unsigned i=0; i<sigma_mean2_.size(); ++i) {
976 : std::string num;
977 2394 : Tools::convert(i,num);
978 4788 : addComponent("sigmaMean-"+num);
979 2394 : componentIsNotPeriodic("sigmaMean-"+num);
980 2394 : valueSigmaMean.push_back(getPntrToComponent("sigmaMean-"+num));
981 4788 : getPntrToComponent("sigmaMean-"+num)->set(std::sqrt(sigma_mean2_[i]));
982 4788 : addComponent("sigma-"+num);
983 2394 : componentIsNotPeriodic("sigma-"+num);
984 2394 : valueSigma.push_back(getPntrToComponent("sigma-"+num));
985 2394 : getPntrToComponent("sigma-"+num)->set(sigma_[i]);
986 2394 : if(noise_type_==GENERIC) {
987 12 : addComponent("ftilde-"+num);
988 6 : componentIsNotPeriodic("ftilde-"+num);
989 6 : valueFtilde.push_back(getPntrToComponent("ftilde-"+num));
990 : }
991 : }
992 15 : } else {
993 34 : addComponent("sigmaMean");
994 17 : componentIsNotPeriodic("sigmaMean");
995 17 : valueSigmaMean.push_back(getPntrToComponent("sigmaMean"));
996 34 : getPntrToComponent("sigmaMean")->set(std::sqrt(sigma_mean2_[0]));
997 34 : addComponent("sigma");
998 17 : componentIsNotPeriodic("sigma");
999 17 : valueSigma.push_back(getPntrToComponent("sigma"));
1000 34 : getPntrToComponent("sigma")->set(sigma_[0]);
1001 : }
1002 :
1003 : // initialize random seed
1004 : unsigned iseed;
1005 32 : if(master) {
1006 24 : auto ts = std::chrono::time_point_cast<std::chrono::nanoseconds>(std::chrono::steady_clock::now()).time_since_epoch().count();
1007 24 : iseed = static_cast<unsigned>(ts)+replica_;
1008 : } else {
1009 8 : iseed = 0;
1010 : }
1011 32 : comm.Sum(&iseed, 1);
1012 : // this is used for ftilde and sigma both the move and the acceptance
1013 : // this is different for each replica
1014 32 : random[0].setSeed(-iseed);
1015 32 : if(doscale_||dooffset_) {
1016 : // in this case we want the same seed everywhere
1017 18 : auto ts = std::chrono::time_point_cast<std::chrono::nanoseconds>(std::chrono::steady_clock::now()).time_since_epoch().count();
1018 18 : iseed = static_cast<unsigned>(ts);
1019 18 : if(master&&nrep_>1) {
1020 6 : multi_sim_comm.Bcast(iseed,0);
1021 : }
1022 18 : comm.Bcast(iseed,0);
1023 : // this is used for scale and offset sampling and acceptance
1024 18 : random[1].setSeed(-iseed);
1025 : }
1026 : // this is used for random chunk of sigmas, and it is different for each replica
1027 32 : if(master) {
1028 24 : auto ts = std::chrono::time_point_cast<std::chrono::nanoseconds>(std::chrono::steady_clock::now()).time_since_epoch().count();
1029 24 : iseed = static_cast<unsigned>(ts)+replica_;
1030 : } else {
1031 8 : iseed = 0;
1032 : }
1033 32 : comm.Sum(&iseed, 1);
1034 32 : random[2].setSeed(-iseed);
1035 :
1036 : // outfile stuff
1037 32 : if(write_stride_>0) {
1038 32 : sfile_.link(*this);
1039 32 : sfile_.open(status_file_name_);
1040 32 : if(fmt_.length()>0) {
1041 1 : sfile_.fmtField(fmt_);
1042 : }
1043 : }
1044 :
1045 64 : log<<" Bibliography "<<plumed.cite("Bonomi, Camilloni, Cavalli, Vendruscolo, Sci. Adv. 2, e150117 (2016)");
1046 32 : if(do_reweight_) {
1047 32 : log<<plumed.cite("Bonomi, Camilloni, Vendruscolo, Sci. Rep. 6, 31232 (2016)");
1048 : }
1049 32 : if(do_optsigmamean_>0) {
1050 0 : log<<plumed.cite("Loehr, Jussupow, Camilloni, J. Chem. Phys. 146, 165102 (2017)");
1051 : }
1052 64 : log<<plumed.cite("Bonomi, Camilloni, Bioinformatics, 33, 3999 (2017)");
1053 32 : log<<"\n";
1054 64 : }
1055 :
1056 64 : Metainference::~Metainference() {
1057 32 : if(sfile_.isOpen()) {
1058 32 : sfile_.close();
1059 : }
1060 192 : }
1061 :
1062 264 : double Metainference::getEnergySP(const std::vector<double> &mean, const std::vector<double> &sigma,
1063 : const double scale, const double offset) {
1064 264 : const double scale2 = scale*scale;
1065 264 : const double sm2 = sigma_mean2_[0];
1066 264 : const double ss2 = sigma[0]*sigma[0] + scale2*sm2;
1067 264 : const double sss = sigma[0]*sigma[0] + sm2;
1068 :
1069 : double ene = 0.0;
1070 264 : #pragma omp parallel num_threads(OpenMP::getNumThreads()) shared(ene)
1071 : {
1072 : #pragma omp for reduction( + : ene)
1073 : for(unsigned i=0; i<narg; ++i) {
1074 : const double dev = scale*mean[i]-parameters[i]+offset;
1075 : const double a2 = 0.5*dev*dev + ss2;
1076 : if(sm2 > 0.0) {
1077 : ene += std::log(2.0*a2/(1.0-std::exp(-a2/sm2)));
1078 : } else {
1079 : ene += std::log(2.0*a2);
1080 : }
1081 : }
1082 : }
1083 : // add one single Jeffrey's prior and one normalisation per data point
1084 264 : ene += 0.5*std::log(sss) + static_cast<double>(narg)*0.5*std::log(0.5*M_PI*M_PI/ss2);
1085 264 : if(doscale_ || doregres_zero_) {
1086 156 : ene += 0.5*std::log(sss);
1087 : }
1088 264 : if(dooffset_) {
1089 0 : ene += 0.5*std::log(sss);
1090 : }
1091 264 : return kbt_ * ene;
1092 : }
1093 :
1094 144 : double Metainference::getEnergySPE(const std::vector<double> &mean, const std::vector<double> &sigma,
1095 : const double scale, const double offset) {
1096 144 : const double scale2 = scale*scale;
1097 : double ene = 0.0;
1098 144 : #pragma omp parallel num_threads(OpenMP::getNumThreads()) shared(ene)
1099 : {
1100 : #pragma omp for reduction( + : ene)
1101 : for(unsigned i=0; i<narg; ++i) {
1102 : const double sm2 = sigma_mean2_[i];
1103 : const double ss2 = sigma[i]*sigma[i] + scale2*sm2;
1104 : const double sss = sigma[i]*sigma[i] + sm2;
1105 : const double dev = scale*mean[i]-parameters[i]+offset;
1106 : const double a2 = 0.5*dev*dev + ss2;
1107 : if(sm2 > 0.0) {
1108 : ene += 0.5*std::log(sss) + 0.5*std::log(0.5*M_PI*M_PI/ss2) + std::log(2.0*a2/(1.0-std::exp(-a2/sm2)));
1109 : } else {
1110 : ene += 0.5*std::log(sss) + 0.5*std::log(0.5*M_PI*M_PI/ss2) + std::log(2.0*a2);
1111 : }
1112 : if(doscale_ || doregres_zero_) {
1113 : ene += 0.5*std::log(sss);
1114 : }
1115 : if(dooffset_) {
1116 : ene += 0.5*std::log(sss);
1117 : }
1118 : }
1119 : }
1120 144 : return kbt_ * ene;
1121 : }
1122 :
1123 144 : double Metainference::getEnergyMIGEN(const std::vector<double> &mean, const std::vector<double> &ftilde, const std::vector<double> &sigma,
1124 : const double scale, const double offset) {
1125 : double ene = 0.0;
1126 144 : #pragma omp parallel num_threads(OpenMP::getNumThreads()) shared(ene)
1127 : {
1128 : #pragma omp for reduction( + : ene)
1129 : for(unsigned i=0; i<narg; ++i) {
1130 : const double inv_sb2 = 1./(sigma[i]*sigma[i]);
1131 : const double inv_sm2 = 1./sigma_mean2_[i];
1132 : double devb = 0;
1133 : if(gen_likelihood_==LIKE_GAUSS) {
1134 : devb = scale*ftilde[i]-parameters[i]+offset;
1135 : } else if(gen_likelihood_==LIKE_LOGN) {
1136 : devb = std::log(scale*ftilde[i]/parameters[i]);
1137 : }
1138 : double devm = mean[i] - ftilde[i];
1139 : // deviation + normalisation + jeffrey
1140 : double normb = 0.;
1141 : if(gen_likelihood_==LIKE_GAUSS) {
1142 : normb = -0.5*std::log(0.5/M_PI*inv_sb2);
1143 : } else if(gen_likelihood_==LIKE_LOGN) {
1144 : normb = -0.5*std::log(0.5/M_PI*inv_sb2/(parameters[i]*parameters[i]));
1145 : }
1146 : const double normm = -0.5*std::log(0.5/M_PI*inv_sm2);
1147 : const double jeffreys = -0.5*std::log(2.*inv_sb2);
1148 : ene += 0.5*devb*devb*inv_sb2 + 0.5*devm*devm*inv_sm2 + normb + normm + jeffreys;
1149 : if(doscale_ || doregres_zero_) {
1150 : ene += jeffreys;
1151 : }
1152 : if(dooffset_) {
1153 : ene += jeffreys;
1154 : }
1155 : }
1156 : }
1157 144 : return kbt_ * ene;
1158 : }
1159 :
1160 108 : double Metainference::getEnergyGJ(const std::vector<double> &mean, const std::vector<double> &sigma,
1161 : const double scale, const double offset) {
1162 108 : const double scale2 = scale*scale;
1163 108 : const double inv_s2 = 1./(sigma[0]*sigma[0] + scale2*sigma_mean2_[0]);
1164 108 : const double inv_sss = 1./(sigma[0]*sigma[0] + sigma_mean2_[0]);
1165 :
1166 : double ene = 0.0;
1167 108 : #pragma omp parallel num_threads(OpenMP::getNumThreads()) shared(ene)
1168 : {
1169 : #pragma omp for reduction( + : ene)
1170 : for(unsigned i=0; i<narg; ++i) {
1171 : double dev = scale*mean[i]-parameters[i]+offset;
1172 : ene += 0.5*dev*dev*inv_s2;
1173 : }
1174 : }
1175 108 : const double normalisation = -0.5*std::log(0.5/M_PI*inv_s2);
1176 108 : const double jeffreys = -0.5*std::log(2.*inv_sss);
1177 : // add Jeffrey's prior in case one sigma for all data points + one normalisation per datapoint
1178 108 : ene += jeffreys + static_cast<double>(narg)*normalisation;
1179 108 : if(doscale_ || doregres_zero_) {
1180 0 : ene += jeffreys;
1181 : }
1182 108 : if(dooffset_) {
1183 108 : ene += jeffreys;
1184 : }
1185 :
1186 108 : return kbt_ * ene;
1187 : }
1188 :
1189 152 : double Metainference::getEnergyGJE(const std::vector<double> &mean, const std::vector<double> &sigma,
1190 : const double scale, const double offset) {
1191 152 : const double scale2 = scale*scale;
1192 :
1193 : double ene = 0.0;
1194 152 : #pragma omp parallel num_threads(OpenMP::getNumThreads()) shared(ene)
1195 : {
1196 : #pragma omp for reduction( + : ene)
1197 : for(unsigned i=0; i<narg; ++i) {
1198 : const double inv_s2 = 1./(sigma[i]*sigma[i] + scale2*sigma_mean2_[i]);
1199 : const double inv_sss = 1./(sigma[i]*sigma[i] + sigma_mean2_[i]);
1200 : double dev = scale*mean[i]-parameters[i]+offset;
1201 : // deviation + normalisation + jeffrey
1202 : const double normalisation = -0.5*std::log(0.5/M_PI*inv_s2);
1203 : const double jeffreys = -0.5*std::log(2.*inv_sss);
1204 : ene += 0.5*dev*dev*inv_s2 + normalisation + jeffreys;
1205 : if(doscale_ || doregres_zero_) {
1206 : ene += jeffreys;
1207 : }
1208 : if(dooffset_) {
1209 : ene += jeffreys;
1210 : }
1211 : }
1212 : }
1213 152 : return kbt_ * ene;
1214 : }
1215 :
1216 36 : void Metainference::moveTilde(const std::vector<double> &mean_, double &old_energy) {
1217 36 : std::vector<double> new_ftilde(sigma_.size());
1218 36 : new_ftilde = ftilde_;
1219 :
1220 : // change all tildes
1221 108 : for(unsigned j=0; j<sigma_.size(); j++) {
1222 72 : const double r3 = random[0].Gaussian();
1223 72 : const double ds3 = Dftilde_*std::sqrt(sigma_mean2_[j])*r3;
1224 72 : new_ftilde[j] = ftilde_[j] + ds3;
1225 : }
1226 : // calculate new energy
1227 36 : double new_energy = getEnergyMIGEN(mean_,new_ftilde,sigma_,scale_,offset_);
1228 :
1229 : // accept or reject
1230 36 : const double delta = ( new_energy - old_energy ) / kbt_;
1231 : // if delta is negative always accept move
1232 36 : if( delta <= 0.0 ) {
1233 36 : old_energy = new_energy;
1234 36 : ftilde_ = new_ftilde;
1235 36 : MCacceptFT_++;
1236 : // otherwise extract random number
1237 : } else {
1238 0 : const double s = random[0].RandU01();
1239 0 : if( s < std::exp(-delta) ) {
1240 0 : old_energy = new_energy;
1241 0 : ftilde_ = new_ftilde;
1242 0 : MCacceptFT_++;
1243 : }
1244 : }
1245 36 : }
1246 :
1247 216 : void Metainference::moveScaleOffset(const std::vector<double> &mean_, double &old_energy) {
1248 216 : double new_scale = scale_;
1249 :
1250 216 : if(doscale_) {
1251 144 : if(scale_prior_==SC_FLAT) {
1252 144 : const double r1 = random[1].Gaussian();
1253 144 : const double ds1 = Dscale_*r1;
1254 144 : new_scale += ds1;
1255 : // check boundaries
1256 144 : if(new_scale > scale_max_) {
1257 0 : new_scale = 2.0 * scale_max_ - new_scale;
1258 : }
1259 144 : if(new_scale < scale_min_) {
1260 0 : new_scale = 2.0 * scale_min_ - new_scale;
1261 : }
1262 : } else {
1263 0 : const double r1 = random[1].Gaussian();
1264 0 : const double ds1 = 0.5*(scale_mu_-new_scale)+Dscale_*std::exp(1)/M_PI*r1;
1265 0 : new_scale += ds1;
1266 : }
1267 : }
1268 :
1269 216 : double new_offset = offset_;
1270 :
1271 216 : if(dooffset_) {
1272 72 : if(offset_prior_==SC_FLAT) {
1273 72 : const double r1 = random[1].Gaussian();
1274 72 : const double ds1 = Doffset_*r1;
1275 72 : new_offset += ds1;
1276 : // check boundaries
1277 72 : if(new_offset > offset_max_) {
1278 0 : new_offset = 2.0 * offset_max_ - new_offset;
1279 : }
1280 72 : if(new_offset < offset_min_) {
1281 0 : new_offset = 2.0 * offset_min_ - new_offset;
1282 : }
1283 : } else {
1284 0 : const double r1 = random[1].Gaussian();
1285 0 : const double ds1 = 0.5*(offset_mu_-new_offset)+Doffset_*std::exp(1)/M_PI*r1;
1286 0 : new_offset += ds1;
1287 : }
1288 : }
1289 :
1290 : // calculate new energy
1291 : double new_energy = 0.;
1292 :
1293 216 : switch(noise_type_) {
1294 36 : case GAUSS:
1295 36 : new_energy = getEnergyGJ(mean_,sigma_,new_scale,new_offset);
1296 : break;
1297 48 : case MGAUSS:
1298 48 : new_energy = getEnergyGJE(mean_,sigma_,new_scale,new_offset);
1299 : break;
1300 48 : case OUTLIERS:
1301 48 : new_energy = getEnergySP(mean_,sigma_,new_scale,new_offset);
1302 : break;
1303 48 : case MOUTLIERS:
1304 48 : new_energy = getEnergySPE(mean_,sigma_,new_scale,new_offset);
1305 : break;
1306 36 : case GENERIC:
1307 36 : new_energy = getEnergyMIGEN(mean_,ftilde_,sigma_,new_scale,new_offset);
1308 : break;
1309 : }
1310 :
1311 : // for the scale/offset we need to consider the total energy
1312 216 : std::vector<double> totenergies(2);
1313 216 : if(master) {
1314 144 : totenergies[0] = old_energy;
1315 144 : totenergies[1] = new_energy;
1316 144 : if(nrep_>1) {
1317 72 : multi_sim_comm.Sum(totenergies);
1318 : }
1319 : } else {
1320 72 : totenergies[0] = 0;
1321 72 : totenergies[1] = 0;
1322 : }
1323 216 : comm.Sum(totenergies);
1324 :
1325 : // accept or reject
1326 216 : const double delta = ( totenergies[1] - totenergies[0] ) / kbt_;
1327 : // if delta is negative always accept move
1328 216 : if( delta <= 0.0 ) {
1329 216 : old_energy = new_energy;
1330 216 : scale_ = new_scale;
1331 216 : offset_ = new_offset;
1332 216 : MCacceptScale_++;
1333 : // otherwise extract random number
1334 : } else {
1335 0 : double s = random[1].RandU01();
1336 0 : if( s < std::exp(-delta) ) {
1337 0 : old_energy = new_energy;
1338 0 : scale_ = new_scale;
1339 0 : offset_ = new_offset;
1340 0 : MCacceptScale_++;
1341 : }
1342 : }
1343 216 : }
1344 :
1345 280 : void Metainference::moveSigmas(const std::vector<double> &mean_, double &old_energy, const unsigned i, const std::vector<unsigned> &indices, bool &breaknow) {
1346 280 : std::vector<double> new_sigma(sigma_.size());
1347 280 : new_sigma = sigma_;
1348 :
1349 : // change MCchunksize_ sigmas
1350 280 : if (MCchunksize_ > 0) {
1351 6 : if ((MCchunksize_ * i) >= sigma_.size()) {
1352 : // This means we are not moving any sigma, so we should break immediately
1353 0 : breaknow = true;
1354 : }
1355 :
1356 : // change random sigmas
1357 12 : for(unsigned j=0; j<MCchunksize_; j++) {
1358 6 : const unsigned shuffle_index = j + MCchunksize_ * i;
1359 6 : if (shuffle_index >= sigma_.size()) {
1360 : // Going any further will segfault but we should still evaluate the sigmas we changed
1361 : break;
1362 : }
1363 6 : const unsigned index = indices[shuffle_index];
1364 6 : const double r2 = random[0].Gaussian();
1365 6 : const double ds2 = Dsigma_[index]*r2;
1366 6 : new_sigma[index] = sigma_[index] + ds2;
1367 : // check boundaries
1368 6 : if(new_sigma[index] > sigma_max_[index]) {
1369 0 : new_sigma[index] = 2.0 * sigma_max_[index] - new_sigma[index];
1370 : }
1371 6 : if(new_sigma[index] < sigma_min_[index]) {
1372 0 : new_sigma[index] = 2.0 * sigma_min_[index] - new_sigma[index];
1373 : }
1374 : }
1375 : } else {
1376 : // change all sigmas
1377 3224 : for(unsigned j=0; j<sigma_.size(); j++) {
1378 2950 : const double r2 = random[0].Gaussian();
1379 2950 : const double ds2 = Dsigma_[j]*r2;
1380 2950 : new_sigma[j] = sigma_[j] + ds2;
1381 : // check boundaries
1382 2950 : if(new_sigma[j] > sigma_max_[j]) {
1383 0 : new_sigma[j] = 2.0 * sigma_max_[j] - new_sigma[j];
1384 : }
1385 2950 : if(new_sigma[j] < sigma_min_[j]) {
1386 0 : new_sigma[j] = 2.0 * sigma_min_[j] - new_sigma[j];
1387 : }
1388 : }
1389 : }
1390 :
1391 280 : if (breaknow) {
1392 : // We didnt move any sigmas, so no sense in evaluating anything
1393 : return;
1394 : }
1395 :
1396 : // calculate new energy
1397 : double new_energy = 0.;
1398 280 : switch(noise_type_) {
1399 36 : case GAUSS:
1400 36 : new_energy = getEnergyGJ(mean_,new_sigma,scale_,offset_);
1401 : break;
1402 52 : case MGAUSS:
1403 52 : new_energy = getEnergyGJE(mean_,new_sigma,scale_,offset_);
1404 : break;
1405 108 : case OUTLIERS:
1406 108 : new_energy = getEnergySP(mean_,new_sigma,scale_,offset_);
1407 : break;
1408 48 : case MOUTLIERS:
1409 48 : new_energy = getEnergySPE(mean_,new_sigma,scale_,offset_);
1410 : break;
1411 36 : case GENERIC:
1412 36 : new_energy = getEnergyMIGEN(mean_,ftilde_,new_sigma,scale_,offset_);
1413 : break;
1414 : }
1415 :
1416 : // accept or reject
1417 280 : const double delta = ( new_energy - old_energy ) / kbt_;
1418 : // if delta is negative always accept move
1419 280 : if( delta <= 0.0 ) {
1420 280 : old_energy = new_energy;
1421 280 : sigma_ = new_sigma;
1422 280 : MCaccept_++;
1423 : // otherwise extract random number
1424 : } else {
1425 0 : const double s = random[0].RandU01();
1426 0 : if( s < std::exp(-delta) ) {
1427 0 : old_energy = new_energy;
1428 0 : sigma_ = new_sigma;
1429 0 : MCaccept_++;
1430 : }
1431 : }
1432 : }
1433 :
1434 280 : double Metainference::doMonteCarlo(const std::vector<double> &mean_) {
1435 : // calculate old energy with the updated coordinates
1436 280 : double old_energy=0.;
1437 :
1438 280 : switch(noise_type_) {
1439 36 : case GAUSS:
1440 36 : old_energy = getEnergyGJ(mean_,sigma_,scale_,offset_);
1441 36 : break;
1442 52 : case MGAUSS:
1443 52 : old_energy = getEnergyGJE(mean_,sigma_,scale_,offset_);
1444 52 : break;
1445 108 : case OUTLIERS:
1446 108 : old_energy = getEnergySP(mean_,sigma_,scale_,offset_);
1447 108 : break;
1448 48 : case MOUTLIERS:
1449 48 : old_energy = getEnergySPE(mean_,sigma_,scale_,offset_);
1450 48 : break;
1451 36 : case GENERIC:
1452 36 : old_energy = getEnergyMIGEN(mean_,ftilde_,sigma_,scale_,offset_);
1453 36 : break;
1454 : }
1455 :
1456 : // do not run MC if this is a replica-exchange trial
1457 280 : if(!getExchangeStep()) {
1458 :
1459 : // Create std::vector of random sigma indices
1460 : std::vector<unsigned> indices;
1461 280 : if (MCchunksize_ > 0) {
1462 12 : for (unsigned j=0; j<sigma_.size(); j++) {
1463 6 : indices.push_back(j);
1464 : }
1465 6 : random[2].Shuffle(indices);
1466 : }
1467 280 : bool breaknow = false;
1468 :
1469 : // cycle on MC steps
1470 560 : for(unsigned i=0; i<MCsteps_; ++i) {
1471 280 : MCtrial_++;
1472 : // propose move for ftilde
1473 280 : if(noise_type_==GENERIC) {
1474 36 : moveTilde(mean_, old_energy);
1475 : }
1476 : // propose move for scale and/or offset
1477 280 : if(doscale_||dooffset_) {
1478 216 : moveScaleOffset(mean_, old_energy);
1479 : }
1480 : // propose move for sigma
1481 280 : moveSigmas(mean_, old_energy, i, indices, breaknow);
1482 : // exit from the loop if this is the case
1483 280 : if(breaknow) {
1484 : break;
1485 : }
1486 : }
1487 :
1488 : /* save the result of the sampling */
1489 : /* ftilde */
1490 280 : if(noise_type_==GENERIC) {
1491 36 : double accept = static_cast<double>(MCacceptFT_) / static_cast<double>(MCtrial_);
1492 36 : valueAcceptFT->set(accept);
1493 108 : for(unsigned i=0; i<sigma_.size(); i++) {
1494 72 : valueFtilde[i]->set(ftilde_[i]);
1495 : }
1496 : }
1497 : /* scale and offset */
1498 280 : if(doscale_ || doregres_zero_) {
1499 150 : valueScale->set(scale_);
1500 : }
1501 280 : if(dooffset_) {
1502 72 : valueOffset->set(offset_);
1503 : }
1504 280 : if(doscale_||dooffset_) {
1505 216 : double accept = static_cast<double>(MCacceptScale_) / static_cast<double>(MCtrial_);
1506 216 : valueAcceptScale->set(accept);
1507 : }
1508 : /* sigmas */
1509 3236 : for(unsigned i=0; i<sigma_.size(); i++) {
1510 2956 : valueSigma[i]->set(sigma_[i]);
1511 : }
1512 280 : double accept = static_cast<double>(MCaccept_) / static_cast<double>(MCtrial_);
1513 280 : valueAccept->set(accept);
1514 : }
1515 :
1516 : // here we sum the score over the replicas to get the full metainference score that we save as a bias
1517 280 : if(master) {
1518 206 : if(nrep_>1) {
1519 74 : multi_sim_comm.Sum(old_energy);
1520 : }
1521 : } else {
1522 74 : old_energy=0;
1523 : }
1524 280 : comm.Sum(old_energy);
1525 :
1526 : // this is the energy with current coordinates and parameters
1527 280 : return old_energy;
1528 : }
1529 :
1530 : /*
1531 : In the following energy-force functions we don't add the normalisation and the jeffreys priors
1532 : because they are not needed for the forces, the correct MetaInference energy is the one calculated
1533 : in the Monte-Carlo
1534 : */
1535 :
1536 108 : void Metainference::getEnergyForceSP(const std::vector<double> &mean, const std::vector<double> &dmean_x,
1537 : const std::vector<double> &dmean_b) {
1538 108 : const double scale2 = scale_*scale_;
1539 108 : const double sm2 = sigma_mean2_[0];
1540 108 : const double ss2 = sigma_[0]*sigma_[0] + scale2*sm2;
1541 108 : std::vector<double> f(narg,0);
1542 :
1543 108 : if(master) {
1544 84 : #pragma omp parallel num_threads(OpenMP::getNumThreads())
1545 : {
1546 : #pragma omp for
1547 : for(unsigned i=0; i<narg; ++i) {
1548 : const double dev = scale_*mean[i]-parameters[i]+offset_;
1549 : const double a2 = 0.5*dev*dev + ss2;
1550 : if(sm2 > 0.0) {
1551 : const double t = std::exp(-a2/sm2);
1552 : const double dt = 1./t;
1553 : const double dit = 1./(1.-dt);
1554 : f[i] = -scale_*dev*(dit/sm2 + 1./a2);
1555 : } else {
1556 : f[i] = -scale_*dev*(1./a2);
1557 : }
1558 : }
1559 : }
1560 : // collect contribution to forces and energy from other replicas
1561 84 : if(nrep_>1) {
1562 24 : multi_sim_comm.Sum(&f[0],narg);
1563 : }
1564 : }
1565 : // intra-replica summation
1566 108 : comm.Sum(&f[0],narg);
1567 :
1568 : double w_tmp = 0.;
1569 720 : for(unsigned i=0; i<narg; ++i) {
1570 612 : setOutputForce(i, kbt_*f[i]*dmean_x[i]);
1571 612 : w_tmp += kbt_*f[i]*dmean_b[i];
1572 : }
1573 :
1574 108 : if(do_reweight_) {
1575 48 : setOutputForce(narg, w_tmp);
1576 96 : getPntrToComponent("biasDer")->set(-w_tmp);
1577 : }
1578 108 : }
1579 :
1580 48 : void Metainference::getEnergyForceSPE(const std::vector<double> &mean, const std::vector<double> &dmean_x,
1581 : const std::vector<double> &dmean_b) {
1582 48 : const double scale2 = scale_*scale_;
1583 48 : std::vector<double> f(narg,0);
1584 :
1585 48 : if(master) {
1586 24 : #pragma omp parallel num_threads(OpenMP::getNumThreads())
1587 : {
1588 : #pragma omp for
1589 : for(unsigned i=0; i<narg; ++i) {
1590 : const double sm2 = sigma_mean2_[i];
1591 : const double ss2 = sigma_[i]*sigma_[i] + scale2*sm2;
1592 : const double dev = scale_*mean[i]-parameters[i]+offset_;
1593 : const double a2 = 0.5*dev*dev + ss2;
1594 : if(sm2 > 0.0) {
1595 : const double t = std::exp(-a2/sm2);
1596 : const double dt = 1./t;
1597 : const double dit = 1./(1.-dt);
1598 : f[i] = -scale_*dev*(dit/sm2 + 1./a2);
1599 : } else {
1600 : f[i] = -scale_*dev*(1./a2);
1601 : }
1602 : }
1603 : }
1604 : // collect contribution to forces and energy from other replicas
1605 24 : if(nrep_>1) {
1606 24 : multi_sim_comm.Sum(&f[0],narg);
1607 : }
1608 : }
1609 48 : comm.Sum(&f[0],narg);
1610 :
1611 : double w_tmp = 0.;
1612 240 : for(unsigned i=0; i<narg; ++i) {
1613 192 : setOutputForce(i, kbt_ * dmean_x[i] * f[i]);
1614 192 : w_tmp += kbt_ * dmean_b[i] *f[i];
1615 : }
1616 :
1617 48 : if(do_reweight_) {
1618 48 : setOutputForce(narg, w_tmp);
1619 96 : getPntrToComponent("biasDer")->set(-w_tmp);
1620 : }
1621 48 : }
1622 :
1623 36 : void Metainference::getEnergyForceGJ(const std::vector<double> &mean, const std::vector<double> &dmean_x,
1624 : const std::vector<double> &dmean_b) {
1625 36 : const double scale2 = scale_*scale_;
1626 36 : double inv_s2=0.;
1627 :
1628 36 : if(master) {
1629 36 : inv_s2 = 1./(sigma_[0]*sigma_[0] + scale2*sigma_mean2_[0]);
1630 36 : if(nrep_>1) {
1631 0 : multi_sim_comm.Sum(inv_s2);
1632 : }
1633 : }
1634 36 : comm.Sum(inv_s2);
1635 :
1636 : double w_tmp = 0.;
1637 36 : #pragma omp parallel num_threads(OpenMP::getNumThreads()) shared(w_tmp)
1638 : {
1639 : #pragma omp for reduction( + : w_tmp)
1640 : for(unsigned i=0; i<narg; ++i) {
1641 : const double dev = scale_*mean[i]-parameters[i]+offset_;
1642 : const double mult = dev*scale_*inv_s2;
1643 : setOutputForce(i, -kbt_*dmean_x[i]*mult);
1644 : w_tmp += kbt_*dmean_b[i]*mult;
1645 : }
1646 : }
1647 :
1648 36 : if(do_reweight_) {
1649 0 : setOutputForce(narg, -w_tmp);
1650 0 : getPntrToComponent("biasDer")->set(w_tmp);
1651 : }
1652 36 : }
1653 :
1654 52 : void Metainference::getEnergyForceGJE(const std::vector<double> &mean, const std::vector<double> &dmean_x,
1655 : const std::vector<double> &dmean_b) {
1656 52 : const double scale2 = scale_*scale_;
1657 52 : std::vector<double> inv_s2(sigma_.size(),0.);
1658 :
1659 52 : if(master) {
1660 1300 : for(unsigned i=0; i<sigma_.size(); ++i) {
1661 1274 : inv_s2[i] = 1./(sigma_[i]*sigma_[i] + scale2*sigma_mean2_[i]);
1662 : }
1663 26 : if(nrep_>1) {
1664 26 : multi_sim_comm.Sum(&inv_s2[0],sigma_.size());
1665 : }
1666 : }
1667 52 : comm.Sum(&inv_s2[0],sigma_.size());
1668 :
1669 : double w_tmp = 0.;
1670 52 : #pragma omp parallel num_threads(OpenMP::getNumThreads()) shared(w_tmp)
1671 : {
1672 : #pragma omp for reduction( + : w_tmp)
1673 : for(unsigned i=0; i<narg; ++i) {
1674 : const double dev = scale_*mean[i]-parameters[i]+offset_;
1675 : const double mult = dev*scale_*inv_s2[i];
1676 : setOutputForce(i, -kbt_*dmean_x[i]*mult);
1677 : w_tmp += kbt_*dmean_b[i]*mult;
1678 : }
1679 : }
1680 :
1681 52 : if(do_reweight_) {
1682 52 : setOutputForce(narg, -w_tmp);
1683 104 : getPntrToComponent("biasDer")->set(w_tmp);
1684 : }
1685 52 : }
1686 :
1687 36 : void Metainference::getEnergyForceMIGEN(const std::vector<double> &mean, const std::vector<double> &dmean_x, const std::vector<double> &dmean_b) {
1688 36 : std::vector<double> inv_s2(sigma_.size(),0.);
1689 36 : std::vector<double> dev(sigma_.size(),0.);
1690 36 : std::vector<double> dev2(sigma_.size(),0.);
1691 :
1692 108 : for(unsigned i=0; i<sigma_.size(); ++i) {
1693 72 : inv_s2[i] = 1./sigma_mean2_[i];
1694 72 : if(master) {
1695 72 : dev[i] = (mean[i]-ftilde_[i]);
1696 72 : dev2[i] = dev[i]*dev[i];
1697 : }
1698 : }
1699 36 : if(master&&nrep_>1) {
1700 0 : multi_sim_comm.Sum(&dev[0],dev.size());
1701 0 : multi_sim_comm.Sum(&dev2[0],dev2.size());
1702 : }
1703 36 : comm.Sum(&dev[0],dev.size());
1704 36 : comm.Sum(&dev2[0],dev2.size());
1705 :
1706 : double dene_b = 0.;
1707 36 : #pragma omp parallel num_threads(OpenMP::getNumThreads()) shared(dene_b)
1708 : {
1709 : #pragma omp for reduction( + : dene_b)
1710 : for(unsigned i=0; i<narg; ++i) {
1711 : const double dene_x = kbt_*inv_s2[i]*dmean_x[i]*dev[i];
1712 : dene_b += kbt_*inv_s2[i]*dmean_b[i]*dev[i];
1713 : setOutputForce(i, -dene_x);
1714 : }
1715 : }
1716 :
1717 36 : if(do_reweight_) {
1718 0 : setOutputForce(narg, -dene_b);
1719 0 : getPntrToComponent("biasDer")->set(dene_b);
1720 : }
1721 36 : }
1722 :
1723 280 : void Metainference::get_weights(const unsigned iselect, double &weight, double &norm, double &neff) {
1724 280 : const double dnrep = static_cast<double>(nrep_);
1725 : // calculate the weights either from BIAS
1726 280 : if(do_reweight_) {
1727 148 : std::vector<double> bias(nrep_,0);
1728 148 : if(master) {
1729 74 : bias[replica_] = getArgument(narg);
1730 74 : if(nrep_>1) {
1731 74 : multi_sim_comm.Sum(&bias[0], nrep_);
1732 : }
1733 : }
1734 148 : comm.Sum(&bias[0], nrep_);
1735 :
1736 : // accumulate weights
1737 148 : const double decay = 1./static_cast<double> (average_weights_stride_);
1738 148 : if(!firstTimeW[iselect]) {
1739 408 : for(unsigned i=0; i<nrep_; ++i) {
1740 272 : const double delta=bias[i]-average_weights_[iselect][i];
1741 272 : average_weights_[iselect][i]+=decay*delta;
1742 : }
1743 : } else {
1744 : firstTimeW[iselect] = false;
1745 36 : for(unsigned i=0; i<nrep_; ++i) {
1746 24 : average_weights_[iselect][i] = bias[i];
1747 : }
1748 : }
1749 :
1750 : // set average back into bias and set norm to one
1751 148 : const double maxbias = *(std::max_element(average_weights_[iselect].begin(), average_weights_[iselect].end()));
1752 444 : for(unsigned i=0; i<nrep_; ++i) {
1753 296 : bias[i] = std::exp((average_weights_[iselect][i]-maxbias)/kbt_);
1754 : }
1755 : // set local weight, norm and weight variance
1756 148 : weight = bias[replica_];
1757 : double w2=0.;
1758 444 : for(unsigned i=0; i<nrep_; ++i) {
1759 296 : w2 += bias[i]*bias[i];
1760 296 : norm += bias[i];
1761 : }
1762 148 : neff = norm*norm/w2;
1763 296 : getPntrToComponent("weight")->set(weight/norm);
1764 : } else {
1765 : // or arithmetic ones
1766 132 : neff = dnrep;
1767 132 : weight = 1.0;
1768 132 : norm = dnrep;
1769 : }
1770 280 : getPntrToComponent("neff")->set(neff);
1771 280 : }
1772 :
1773 280 : void Metainference::get_sigma_mean(const unsigned iselect, const double weight, const double norm, const double neff, const std::vector<double> &mean) {
1774 280 : const double dnrep = static_cast<double>(nrep_);
1775 280 : std::vector<double> sigma_mean2_tmp(sigma_mean2_.size(), 0.);
1776 :
1777 280 : if(do_optsigmamean_>0) {
1778 : // remove first entry of the history std::vector
1779 0 : if(sigma_mean2_last_[iselect][0].size()==optsigmamean_stride_&&optsigmamean_stride_>0)
1780 0 : for(unsigned i=0; i<narg; ++i) {
1781 0 : sigma_mean2_last_[iselect][i].erase(sigma_mean2_last_[iselect][i].begin());
1782 : }
1783 : /* this is the current estimate of sigma mean for each argument
1784 : there is one of this per argument in any case because it is
1785 : the maximum among these to be used in case of GAUSS/OUTLIER */
1786 0 : std::vector<double> sigma_mean2_now(narg,0);
1787 0 : if(master) {
1788 0 : for(unsigned i=0; i<narg; ++i) {
1789 0 : sigma_mean2_now[i] = weight*(getArgument(i)-mean[i])*(getArgument(i)-mean[i]);
1790 : }
1791 0 : if(nrep_>1) {
1792 0 : multi_sim_comm.Sum(&sigma_mean2_now[0], narg);
1793 : }
1794 : }
1795 0 : comm.Sum(&sigma_mean2_now[0], narg);
1796 0 : for(unsigned i=0; i<narg; ++i) {
1797 0 : sigma_mean2_now[i] *= 1.0/(neff-1.)/norm;
1798 : }
1799 :
1800 : // add sigma_mean2 to history
1801 0 : if(optsigmamean_stride_>0) {
1802 0 : for(unsigned i=0; i<narg; ++i) {
1803 0 : sigma_mean2_last_[iselect][i].push_back(sigma_mean2_now[i]);
1804 : }
1805 : } else {
1806 0 : for(unsigned i=0; i<narg; ++i)
1807 0 : if(sigma_mean2_now[i] > sigma_mean2_last_[iselect][i][0]) {
1808 0 : sigma_mean2_last_[iselect][i][0] = sigma_mean2_now[i];
1809 : }
1810 : }
1811 :
1812 0 : if(noise_type_==MGAUSS||noise_type_==MOUTLIERS||noise_type_==GENERIC) {
1813 0 : for(unsigned i=0; i<narg; ++i) {
1814 : /* set to the maximum in history std::vector */
1815 0 : sigma_mean2_tmp[i] = *max_element(sigma_mean2_last_[iselect][i].begin(), sigma_mean2_last_[iselect][i].end());
1816 : /* the standard error of the mean */
1817 0 : valueSigmaMean[i]->set(std::sqrt(sigma_mean2_tmp[i]));
1818 0 : if(noise_type_==GENERIC) {
1819 0 : sigma_min_[i] = std::sqrt(sigma_mean2_tmp[i]);
1820 0 : if(sigma_[i] < sigma_min_[i]) {
1821 0 : sigma_[i] = sigma_min_[i];
1822 : }
1823 : }
1824 : }
1825 0 : } else if(noise_type_==GAUSS||noise_type_==OUTLIERS) {
1826 : // find maximum for each data point
1827 : std::vector <double> max_values;
1828 0 : for(unsigned i=0; i<narg; ++i) {
1829 0 : max_values.push_back(*max_element(sigma_mean2_last_[iselect][i].begin(), sigma_mean2_last_[iselect][i].end()));
1830 : }
1831 : // find maximum across data points
1832 0 : const double max_now = *max_element(max_values.begin(), max_values.end());
1833 : // set new value
1834 0 : sigma_mean2_tmp[0] = max_now;
1835 0 : valueSigmaMean[0]->set(std::sqrt(sigma_mean2_tmp[0]));
1836 : }
1837 : // endif sigma mean optimization
1838 : // start sigma max optimization
1839 0 : if(do_optsigmamean_>1&&!sigmamax_opt_done_) {
1840 0 : for(unsigned i=0; i<sigma_max_.size(); i++) {
1841 0 : if(sigma_max_est_[i]<sigma_mean2_tmp[i]&&optimized_step_>optsigmamean_stride_) {
1842 0 : sigma_max_est_[i]=sigma_mean2_tmp[i];
1843 : }
1844 : // ready to set once and for all the value of sigma_max
1845 0 : if(optimized_step_==N_optimized_step_) {
1846 0 : sigmamax_opt_done_=true;
1847 0 : for(unsigned ii=0; ii<sigma_max_.size(); ii++) {
1848 0 : sigma_max_[ii]=std::sqrt(sigma_max_est_[ii]*dnrep);
1849 0 : Dsigma_[ii] = 0.05*(sigma_max_[ii] - sigma_min_[ii]);
1850 0 : if(sigma_[ii]>sigma_max_[ii]) {
1851 0 : sigma_[ii]=sigma_max_[ii];
1852 : }
1853 : }
1854 : }
1855 : }
1856 0 : optimized_step_++;
1857 : }
1858 : // end sigma max optimization
1859 : } else {
1860 280 : if(noise_type_==MGAUSS||noise_type_==MOUTLIERS||noise_type_==GENERIC) {
1861 2948 : for(unsigned i=0; i<narg; ++i) {
1862 2812 : sigma_mean2_tmp[i] = sigma_mean2_last_[iselect][i][0];
1863 2812 : valueSigmaMean[i]->set(std::sqrt(sigma_mean2_tmp[i]));
1864 : }
1865 144 : } else if(noise_type_==GAUSS||noise_type_==OUTLIERS) {
1866 144 : sigma_mean2_tmp[0] = sigma_mean2_last_[iselect][0][0];
1867 144 : valueSigmaMean[0]->set(std::sqrt(sigma_mean2_tmp[0]));
1868 : }
1869 : }
1870 :
1871 280 : sigma_mean2_ = sigma_mean2_tmp;
1872 280 : }
1873 :
1874 280 : void Metainference::replica_averaging(const double weight, const double norm, std::vector<double> &mean, std::vector<double> &dmean_b) {
1875 280 : if(master) {
1876 2236 : for(unsigned i=0; i<narg; ++i) {
1877 2030 : mean[i] = weight/norm*getArgument(i);
1878 : }
1879 206 : if(nrep_>1) {
1880 74 : multi_sim_comm.Sum(&mean[0], narg);
1881 : }
1882 : }
1883 280 : comm.Sum(&mean[0], narg);
1884 : // set the derivative of the mean with respect to the bias
1885 3776 : for(unsigned i=0; i<narg; ++i) {
1886 3496 : dmean_b[i] = weight/norm/kbt_*(getArgument(i)-mean[i])/static_cast<double>(average_weights_stride_);
1887 : }
1888 :
1889 : // this is only for generic metainference
1890 280 : if(firstTime) {
1891 28 : ftilde_ = mean;
1892 28 : firstTime = false;
1893 : }
1894 280 : }
1895 :
1896 6 : void Metainference::do_regression_zero(const std::vector<double> &mean) {
1897 : // parameters[i] = scale_ * mean[i]: find scale_ with linear regression
1898 : double num = 0.0;
1899 : double den = 0.0;
1900 48 : for(unsigned i=0; i<parameters.size(); ++i) {
1901 42 : num += mean[i] * parameters[i];
1902 42 : den += mean[i] * mean[i];
1903 : }
1904 6 : if(den>0) {
1905 6 : scale_ = num / den;
1906 : } else {
1907 0 : scale_ = 1.0;
1908 : }
1909 6 : }
1910 :
1911 280 : void Metainference::calculate() {
1912 : // get step
1913 280 : const long long int step = getStep();
1914 :
1915 : unsigned iselect = 0;
1916 : // set the value of selector for REM-like stuff
1917 280 : if(selector_.length()>0) {
1918 0 : iselect = static_cast<unsigned>(plumed.passMap[selector_]);
1919 : }
1920 :
1921 : /* 1) collect weights */
1922 280 : double weight = 0.;
1923 280 : double neff = 0.;
1924 280 : double norm = 0.;
1925 280 : get_weights(iselect, weight, norm, neff);
1926 :
1927 : /* 2) calculate average */
1928 280 : std::vector<double> mean(narg,0);
1929 : // this is the derivative of the mean with respect to the argument
1930 280 : std::vector<double> dmean_x(narg,weight/norm);
1931 : // this is the derivative of the mean with respect to the bias
1932 280 : std::vector<double> dmean_b(narg,0);
1933 : // calculate it
1934 280 : replica_averaging(weight, norm, mean, dmean_b);
1935 :
1936 : /* 3) calculates parameters */
1937 280 : get_sigma_mean(iselect, weight, norm, neff, mean);
1938 :
1939 : // in case of regression with zero intercept, calculate scale
1940 280 : if(doregres_zero_ && step%nregres_zero_==0) {
1941 6 : do_regression_zero(mean);
1942 : }
1943 :
1944 : /* 4) run monte carlo */
1945 280 : double ene = doMonteCarlo(mean);
1946 :
1947 : // calculate bias and forces
1948 280 : switch(noise_type_) {
1949 36 : case GAUSS:
1950 36 : getEnergyForceGJ(mean, dmean_x, dmean_b);
1951 : break;
1952 52 : case MGAUSS:
1953 52 : getEnergyForceGJE(mean, dmean_x, dmean_b);
1954 : break;
1955 108 : case OUTLIERS:
1956 108 : getEnergyForceSP(mean, dmean_x, dmean_b);
1957 : break;
1958 48 : case MOUTLIERS:
1959 48 : getEnergyForceSPE(mean, dmean_x, dmean_b);
1960 : break;
1961 36 : case GENERIC:
1962 36 : getEnergyForceMIGEN(mean, dmean_x, dmean_b);
1963 : break;
1964 : }
1965 :
1966 280 : setBias(ene);
1967 280 : }
1968 :
1969 32 : void Metainference::writeStatus() {
1970 32 : sfile_.rewind();
1971 32 : sfile_.printField("time",getTimeStep()*getStep());
1972 : //nsel
1973 64 : for(unsigned i=0; i<sigma_mean2_last_.size(); i++) {
1974 : std::string msg_i,msg_j;
1975 32 : Tools::convert(i,msg_i);
1976 : std::vector <double> max_values;
1977 : //narg
1978 2518 : for(unsigned j=0; j<narg; ++j) {
1979 2486 : Tools::convert(j,msg_j);
1980 2486 : std::string msg = msg_i+"_"+msg_j;
1981 2486 : if(noise_type_==MGAUSS||noise_type_==MOUTLIERS||noise_type_==GENERIC) {
1982 7182 : sfile_.printField("sigmaMean_"+msg,std::sqrt(*max_element(sigma_mean2_last_[i][j].begin(), sigma_mean2_last_[i][j].end())));
1983 : } else {
1984 : // find maximum for each data point
1985 184 : max_values.push_back(*max_element(sigma_mean2_last_[i][j].begin(), sigma_mean2_last_[i][j].end()));
1986 : }
1987 : }
1988 32 : if(noise_type_==GAUSS||noise_type_==OUTLIERS) {
1989 : // find maximum across data points
1990 17 : const double max_now = std::sqrt(*max_element(max_values.begin(), max_values.end()));
1991 17 : Tools::convert(0,msg_j);
1992 17 : std::string msg = msg_i+"_"+msg_j;
1993 34 : sfile_.printField("sigmaMean_"+msg, max_now);
1994 : }
1995 : }
1996 2443 : for(unsigned i=0; i<sigma_.size(); ++i) {
1997 : std::string msg;
1998 2411 : Tools::convert(i,msg);
1999 4822 : sfile_.printField("sigma_"+msg,sigma_[i]);
2000 : }
2001 2443 : for(unsigned i=0; i<sigma_max_.size(); ++i) {
2002 : std::string msg;
2003 2411 : Tools::convert(i,msg);
2004 4822 : sfile_.printField("sigma_max_"+msg,sigma_max_[i]);
2005 : }
2006 32 : if(noise_type_==GENERIC) {
2007 9 : for(unsigned i=0; i<ftilde_.size(); ++i) {
2008 : std::string msg;
2009 6 : Tools::convert(i,msg);
2010 12 : sfile_.printField("ftilde_"+msg,ftilde_[i]);
2011 : }
2012 : }
2013 32 : sfile_.printField("scale0_",scale_);
2014 32 : sfile_.printField("offset0_",offset_);
2015 64 : for(unsigned i=0; i<average_weights_.size(); i++) {
2016 : std::string msg_i;
2017 32 : Tools::convert(i,msg_i);
2018 64 : sfile_.printField("weight_"+msg_i,average_weights_[i][replica_]);
2019 : }
2020 32 : sfile_.printField();
2021 32 : sfile_.flush();
2022 32 : }
2023 :
2024 280 : void Metainference::update() {
2025 : // write status file
2026 280 : if(write_stride_>0&& (getStep()%write_stride_==0 || getCPT()) ) {
2027 32 : writeStatus();
2028 : }
2029 280 : }
2030 :
2031 : }
2032 : }
2033 :
|
