1 /* +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
2 Copyright (c) 2012-2021 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 #include "HistogramBead.h"
23 #include <vector>
24 #include <limits>
25 #include "Tools.h"
26 #include "Keywords.h"
27
28 namespace PLMD {
29
30 //+PLUMEDOC INTERNAL histogrambead
31 /*
32 A function that can be used to calculate whether quantities are between fixed upper and lower bounds.
33
34 If we have multiple instances of a variable we can estimate the probability density function
35 for that variable using a process called kernel density estimation:
36
37 \f[
38 P(s) = \sum_i K\left( \frac{s - s_i}{w} \right)
39 \f]
40
41 In this equation \f$K\f$ is a symmetric function that must integrate to one that is often
42 called a kernel function and \f$w\f$ is a smearing parameter. From a probability density function calculated using
43 kernel density estimation we can calculate the number/fraction of values between an upper and lower
44 bound using:
45
46 \f[
47 w(s) = \int_a^b \sum_i K\left( \frac{s - s_i}{w} \right)
48 \f]
49
50 All the input to calculate a quantity like \f$w(s)\f$ is generally provided through a single
51 keyword that will have the following form:
52
53 KEYWORD={TYPE UPPER=\f$a\f$ LOWER=\f$b\f$ SMEAR=\f$\frac{w}{b-a}\f$}
54
55 This will calculate the number of values between \f$a\f$ and \f$b\f$. To calculate
56 the fraction of values you add the word NORM to the input specification. If the
57 function keyword SMEAR is not present \f$w\f$ is set equal to \f$0.5(b-a)\f$. Finally,
58 type should specify one of the kernel types that is present in plumed. These are listed
59 in the table below:
60
61 <table align=center frame=void width=95%% cellpadding=5%%>
62 <tr>
63 <td> TYPE </td> <td> FUNCTION </td>
64 </tr> <tr>
65 <td> GAUSSIAN </td> <td> \f$\frac{1}{\sqrt{2\pi}w} \exp\left( -\frac{(s-s_i)^2}{2w^2} \right)\f$ </td>
66 </tr> <tr>
67 <td> TRIANGULAR </td> <td> \f$ \frac{1}{2w} \left( 1. - \left| \frac{s-s_i}{w} \right| \right) \quad \frac{s-s_i}{w}<1 \f$ </td>
68 </tr>
69 </table>
70
71 Some keywords can also be used to calculate a discrete version of the histogram. That
72 is to say the number of values between \f$a\f$ and \f$b\f$, the number of values between
73 \f$b\f$ and \f$c\f$ and so on. A keyword that specifies this sort of calculation would look
74 something like
75
76 KEYWORD={TYPE UPPER=\f$a\f$ LOWER=\f$b\f$ NBINS=\f$n\f$ SMEAR=\f$\frac{w}{n(b-a)}\f$}
77
78 This specification would calculate the following vector of quantities:
79
80 \f[
81 w_j(s) = \int_{a + \frac{j-1}{n}(b-a)}^{a + \frac{j}{n}(b-a)} \sum_i K\left( \frac{s - s_i}{w} \right)
82 \f]
83
84 */
85 //+ENDPLUMEDOC
86
registerKeywords(Keywords & keys)87 void HistogramBead::registerKeywords( Keywords& keys ) {
88 keys.add("compulsory","LOWER","the lower boundary for this particular bin");
89 keys.add("compulsory","UPPER","the upper boundary for this particular bin");
90 keys.add("compulsory","SMEAR","0.5","the amount to smear the Gaussian for each value in the distribution");
91 }
92
HistogramBead()93 HistogramBead::HistogramBead():
94 init(false),
95 lowb(0.0),
96 highb(0.0),
97 width(0.0),
98 cutoff(std::numeric_limits<double>::max()),
99 type(gaussian),
100 periodicity(unset),
101 min(0.0),
102 max(0.0),
103 max_minus_min(0.0),
104 inv_max_minus_min(0.0)
105 {
106 }
107
description() const108 std::string HistogramBead::description() const {
109 std::ostringstream ostr;
110 ostr<<"between "<<lowb<<" and "<<highb<<" width of gaussian window equals "<<width;
111 return ostr.str();
112 }
113
generateBins(const std::string & params,std::vector<std::string> & bins)114 void HistogramBead::generateBins( const std::string& params, std::vector<std::string>& bins ) {
115 std::vector<std::string> data=Tools::getWords(params);
116 plumed_massert(data.size()>=1,"There is no input for this keyword");
117
118 std::string name=data[0];
119
120 unsigned nbins; std::vector<double> range(2); std::string smear;
121 bool found_nb=Tools::parse(data,"NBINS",nbins);
122 plumed_massert(found_nb,"Number of bins in histogram not found");
123 bool found_r=Tools::parse(data,"LOWER",range[0]);
124 plumed_massert(found_r,"Lower bound for histogram not specified");
125 found_r=Tools::parse(data,"UPPER",range[1]);
126 plumed_massert(found_r,"Upper bound for histogram not specified");
127 plumed_massert(range[0]<range[1],"Range specification is dubious");
128 bool found_b=Tools::parse(data,"SMEAR",smear);
129 if(!found_b) { Tools::convert(0.5,smear); }
130
131 std::string lb,ub; double delr = ( range[1]-range[0] ) / static_cast<double>( nbins );
132 for(unsigned i=0; i<nbins; ++i) {
133 Tools::convert( range[0]+i*delr, lb );
134 Tools::convert( range[0]+(i+1)*delr, ub );
135 bins.push_back( name + " " + "LOWER=" + lb + " " + "UPPER=" + ub + " " + "SMEAR=" + smear );
136 }
137 plumed_assert(bins.size()==nbins);
138 }
139
set(const std::string & params,std::string & errormsg)140 void HistogramBead::set( const std::string& params, std::string& errormsg ) {
141 std::vector<std::string> data=Tools::getWords(params);
142 if(data.size()<1) {
143 errormsg="No input has been specified";
144 return;
145 }
146
147 std::string name=data[0]; const double DP2CUTOFF=6.25;
148 if(name=="GAUSSIAN") { type=gaussian; cutoff=sqrt(2.0*DP2CUTOFF); }
149 else if(name=="TRIANGULAR") { type=triangular; cutoff=1.; }
150 else plumed_merror("cannot understand kernel type " + name );
151
152 double smear;
153 bool found_r=Tools::parse(data,"LOWER",lowb);
154 if( !found_r ) errormsg="Lower bound has not been specified use LOWER";
155 found_r=Tools::parse(data,"UPPER",highb);
156 if( !found_r ) errormsg="Upper bound has not been specified use UPPER";
157 if( lowb>=highb ) errormsg="Lower bound is higher than upper bound";
158
159 smear=0.5; Tools::parse(data,"SMEAR",smear);
160 width=smear*(highb-lowb); init=true;
161 }
162
set(double l,double h,double w)163 void HistogramBead::set( double l, double h, double w) {
164 init=true; lowb=l; highb=h; width=w;
165 const double DP2CUTOFF=6.25;
166 if( type==gaussian ) cutoff=sqrt(2.0*DP2CUTOFF);
167 else if( type==triangular ) cutoff=1.;
168 else plumed_error();
169 }
170
setKernelType(const std::string & ktype)171 void HistogramBead::setKernelType( const std::string& ktype ) {
172 if(ktype=="gaussian") type=gaussian;
173 else if(ktype=="triangular") type=triangular;
174 else plumed_merror("cannot understand kernel type " + ktype );
175 }
176
calculate(double x,double & df) const177 double HistogramBead::calculate( double x, double& df ) const {
178 plumed_dbg_assert(init && periodicity!=unset );
179 double lowB, upperB, f;
180 if( type==gaussian ) {
181 lowB = difference( x, lowb ) / ( sqrt(2.0) * width );
182 upperB = difference( x, highb ) / ( sqrt(2.0) * width );
183 df = ( exp( -lowB*lowB ) - exp( -upperB*upperB ) ) / ( sqrt(2*pi)*width );
184 f = 0.5*( erf( upperB ) - erf( lowB ) );
185 } else if( type==triangular ) {
186 lowB = ( difference( x, lowb ) / width );
187 upperB = ( difference( x, highb ) / width );
188 df=0;
189 if( fabs(lowB)<1. ) df = (1 - fabs(lowB)) / width;
190 if( fabs(upperB)<1. ) df -= (1 - fabs(upperB)) / width;
191 if (upperB<=-1. || lowB >=1.) {
192 f=0.;
193 } else {
194 double ia, ib;
195 if( lowB>-1.0 ) { ia=lowB; } else { ia=-1.0; }
196 if( upperB<1.0 ) { ib=upperB; } else { ib=1.0; }
197 f = (ib*(2.-fabs(ib))-ia*(2.-fabs(ia)))*0.5;
198 }
199 } else {
200 plumed_merror("function type does not exist");
201 }
202 return f;
203 }
204
calculateWithCutoff(double x,double & df) const205 double HistogramBead::calculateWithCutoff( double x, double& df ) const {
206 plumed_dbg_assert(init && periodicity!=unset );
207
208 double lowB, upperB, f;
209 lowB = difference( x, lowb ) / width ; upperB = difference( x, highb ) / width;
210 if( upperB<=-cutoff || lowB>=cutoff ) { df=0; return 0; }
211
212 if( type==gaussian ) {
213 lowB /= sqrt(2.0); upperB /= sqrt(2.0);
214 df = ( exp( -lowB*lowB ) - exp( -upperB*upperB ) ) / ( sqrt(2*pi)*width );
215 f = 0.5*( erf( upperB ) - erf( lowB ) );
216 } else if( type==triangular ) {
217 df=0;
218 if( fabs(lowB)<1. ) df = (1 - fabs(lowB)) / width;
219 if( fabs(upperB)<1. ) df -= (1 - fabs(upperB)) / width;
220 if (upperB<=-1. || lowB >=1.) {
221 f=0.;
222 } else {
223 double ia, ib;
224 if( lowB>-1.0 ) { ia=lowB; } else { ia=-1.0; }
225 if( upperB<1.0 ) { ib=upperB; } else { ib=1.0; }
226 f = (ib*(2.-fabs(ib))-ia*(2.-fabs(ia)))*0.5;
227 }
228 } else {
229 plumed_merror("function type does not exist");
230 }
231 return f;
232 }
233
lboundDerivative(const double & x) const234 double HistogramBead::lboundDerivative( const double& x ) const {
235 if( type==gaussian ) {
236 double lowB = difference( x, lowb ) / ( sqrt(2.0) * width );
237 return exp( -lowB*lowB ) / ( sqrt(2*pi)*width );
238 } else if ( type==triangular ) {
239 plumed_error();
240 // lowB = fabs( difference( x, lowb ) / width );
241 // if( lowB<1 ) return ( 1 - (lowB) ) / 2*width;
242 // else return 0;
243 } else {
244 plumed_merror("function type does not exist");
245 }
246 return 0;
247 }
248
uboundDerivative(const double & x) const249 double HistogramBead::uboundDerivative( const double& x ) const {
250 plumed_dbg_assert(init && periodicity!=unset );
251 if( type==gaussian ) {
252 double upperB = difference( x, highb ) / ( sqrt(2.0) * width );
253 return exp( -upperB*upperB ) / ( sqrt(2*pi)*width );
254 } else if ( type==triangular ) {
255 plumed_error();
256 // upperB = fabs( difference( x, highb ) / width );
257 // if( upperB<1 ) return ( 1 - (upperB) ) / 2*width;
258 // else return 0;
259 } else {
260 plumed_merror("function type does not exist");
261 }
262 return 0;
263 }
264
265 }
266