def build_GaussianProcess_target_code(self):
with open('evaluate_simulation.m','w') as fh:
writeline('function mean_overlap = evaluate_simulation(iteration, nres)', fh)
writeline(' ', fh)
writeline('%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%', fh)
writeline('% ', fh)
writeline('%% ', fh)
writeline('%% THIS CODE WAS DYNAMICALLY GENERATED BY CAMeLOT ', fh)
writeline('%% Code generated at ' + str(time.strftime("%c")), fh)
writeline('%% Generated by CAMeLOT version ' + self.MolTemplateParameters['CAMELOT_VERSION'], fh)
writeline('%% ', fh)
writeline('%% cg_opt - Optimization framework script which runs the BayesOpt package', fh)
writeline('%% to perform Gaussian Process Bayesian optimization', fh)
writeline('%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%', fh)
writeline(' ', fh)
writeline('% source the dcd reading functionality', fh)
writeline('addpath %s '%(configs.MATDCD_LOCATION), fh)
writeline('% box length', fh)
writeline('bl = [255.1, 255.1, 255.1];', fh)
writeline(' ', fh)
writeline('% discard the first 20% as equlibration', fh)
writeline('equil = %i ;'%(self.SIM_EQUIL_FRACTION*(self.SIM_NSTEPS/self.SIM_DCD_OUT)), fh)
writeline('binwidth=%3.3f;'%(self.RES_RES_BIN_SIZE), fh)
writeline('binhalf=binwidth/2;', fh)
writeline(' ', fh)
writeline('% remove the first 1e5 steps', fh)
writeline('equil = 101;', fh)
writeline(' ', fh)
writeline('% get bead positions', fh)
writeline("da = readdcd([num2str(iteration) '/optimization.dcd'],1:nres);",fh)
writeline('xyzmat=da(equil:end,:);', fh)
writeline(' ', fh)
writeline('for i=1:nres-1', fh)
writeline(' for j=1+1:nres', fh)
writeline(' positions_i = xyzmat(:, 1+3*(i-1):3+3*(i-1));', fh)
writeline(' positions_j = xyzmat(:, 1+3*(j-1):3+3*(j-1));', fh)
writeline(' ', fh)
writeline(' dPos = positions_i - positions_j;', fh)
writeline(' % correct for periodicity in x/y/z', fh)
writeline(' xcorrect=find(abs(dPos(:,1))>bl(1)*0.5);', fh)
writeline(' dPos(xcorrect,1)=bl(1)-abs(dPos(xcorrect,1));', fh)
writeline(' ', fh)
writeline(' ycorrect=find(abs(dPos(:,2))>bl(2)*0.5);', fh)
writeline(' dPos(ycorrect,2)=bl(2)-abs(dPos(ycorrect,2));', fh)
writeline(' ', fh)
writeline(' zcorrect=find(abs(dPos(:,3))>bl(3)*0.5);', fh)
writeline(' dPos(zcorrect,3)=bl(3)-abs(dPos(zcorrect,3));', fh)
writeline(' ', fh)
writeline(' % determine full set of distance value', fh)
writeline(' raw_distance{i,j} = sqrt(sum(dPos.^2,2));', fh)
writeline(' ', fh)
writeline(" % note we have to offset the bin edges because numpy and matlab have different", fh)
writeline(" % interepetations of how the edges should be used...", fh)
writeline(" vals = histc(raw_distance{i,j},%3.3f:%3.3f:%3.3f-(2*binwidth))';" % (self.RES_RES_BIN_START, self.RES_RES_BIN_SIZE, self.RES_RES_BIN_END), fh)
writeline(' vals = vals./sum(vals);', fh)
writeline(' ', fh)
writeline(' CG_histogram{i,j} = vals;', fh)
writeline(' end', fh)
writeline('end', fh)
writeline(' ', fh)
writeline(' ', fh)
writeline('% extract pre-computed all atom distance histograms', fh)
writeline("AA_distance_distributions=load('%s');"%(self.RES_RES_DISTANCE_FILE), fh)
writeline(' ', fh)
writeline('for i = 1:size(AA_distance_distributions,1)', fh)
writeline(' line = AA_distance_distributions(i,:);', fh)
writeline(' res1 = line(1);', fh)
writeline(' res2 = line(2);', fh)
writeline(' AA_histogram{res1,res2} = line(3:end);', fh)
writeline('end', fh)
writeline(' ', fh)
writeline('% compute all overlaps', fh)
writeline('for i=1:nres-1', fh)
writeline(' for j=i+1:nres', fh)
writeline(' overlap(i,j) = (2-sum(abs(AA_histogram{i,j} - CG_histogram{i,j})))/2;', fh)
writeline(' end', fh)
writeline('end', fh)
writeline(' ', fh)
writeline('% create a vector of the relevant overlaps as defined by the minimium', fh)
writeline('% separation of interest (minsep)', fh)
writeline('minsep=5;', fh)
writeline('relevant = [];', fh)
writeline('for i=1:nres-minsep', fh)
writeline(' for j=i+minsep:nres', fh)
writeline(' relevant = [relevant; (2-sum(abs(AA_histogram{i,j} - CG_histogram{i,j})))/2];', fh)
writeline(' end', fh)
writeline('end',fh)
writeline(' ',fh)
writeline('mean_overlap = 1 - mean(relevant);', fh)
writeline(' ', fh)
writeline(' ', fh)
writeline(' ', fh)
writeline(' ', fh)
def build_nvt_simulation_code(self):
"""
Function which constructs custom code which itself constructs a LAMMPS configuration file
needed to run a production NVT simulation.
Right now (0.1.0) we don't allow customization of the NVT configuration file, but the whole
point of creating a function which writes code which writes a configuration file is that
you can configure that second codeset to include custom variables as defined during the
initial setup.
The point is, the 'generate_nvt.py' file which ends up getting created is a stand alone
custom file which will build a system-bespoke LAMMPS configuration file as a totally
independent script.
This means we can use it as part of the MATLAB drive GPBO procedure, where the MATLAB
code calls these generate_nvt.py script, and keep the idea of creating a system specific
simulation set for optimization.
"""
with open('generate_nvt.py', 'w') as fh:
writeline('import random', fh)
writeline('import time', fh)
writeline('#<><><><><><><><><><><><><><><><><><><><><><><><><><><><><>', fh)
writeline('# ', fh)
writeline('# ', fh)
writeline('# THIS CODE WAS DYNAMICALLY GENERATED BY CAMeLOT ', fh)
writeline('# Code generated at ' + str(time.strftime("%c")), fh)
writeline('# Generated by CAMeLOT version ' + self.MolTemplateParameters['CAMELOT_VERSION'] + "", fh)
writeline('#', fh)
writeline('#', fh)
writeline('# generate_nvt.py', fh)
writeline('# This code is to be executed in the main simulation root subdirectory, and will', fh)
writeline('# generate an NVT production simulation LAMMPS configuration file', fh)
writeline('# with a unique and random SEED.', fh)
writeline('# That config. file should then be moved into an iteration subdirectory', fh)
writeline('# where the simulation should be run', fh)
writeline('#<><><><><><><><><><><><><><><><><><><><><><><><><><><><><>', fh)
writeline('', fh)
writeline('# >>>>>>>>>>>>>>>>>>>>>', fh)
writeline('# First up we need to define the Lagevin string...', fh)
writeline('', fh)
writeline("# Seed the PRNG using the OS' entropy pool", fh)
writeline("random.seed()", fh)
writeline('', fh)
writeline("# First read in the dampening parameters and construct", fh)
writeline("# a custom scale line", fh)
writeline("with open('%s', 'r') as fh:" % self.DAMP_PARAMS, fh)
writeline(" DPs = fh.readlines()", fh)
writeline("DP = DPs[0].strip()", fh)
writeline(' ', fh)
writeline("TEMP = %3.3f"%self.TEMP, fh)
writeline("RAND = random.randint(0,10000)", fh)
writeline("dampline = 'fix fxlan all langevin ' + str(TEMP) + ' ' + str(TEMP) + ' 2000.0 ' + str(RAND) + ' ' + DP", fh)
writeline("with open('run.in.nvt', 'w') as fh:", fh)
writeline(" fh.write('# -------- CAMeLOT NVT simulation: ---------\n')" , fh)
writeline(" fh.write('# This file defines a CAMeLOT simulation file,\n')", fh)
writeline(" fh.write('# and was generated by the generate_nvt.py script\n')", fh)
writeline(" fh.write('# at ' + str(time.strftime('%c')) +'\n')", fh)
writeline(" fh.write('\n')", fh)
writeline(" fh.write('# -- Init section --\n')", fh)
writeline(" fh.write('include system.in.init\n')", fh)
writeline(" fh.write('\n')", fh)
writeline(" fh.write('# -- Atom definition section -- \n')", fh)
writeline(" fh.write('read_restart system_after_min.rst\n')", fh)
writeline(" fh.write('\n')", fh)
writeline(" fh.write('# -- Settings Section --\n') ", fh)
writeline(" fh.write('include system.in.settings\n')", fh)
writeline(" fh.write('\n')", fh)
writeline(" fh.write('# -- Run section --\n')", fh)
writeline(" fh.write('timestep 2.0\n')", fh)
writeline(" fh.write('dump 1 all dcd %i optimization.dcd\n')" % self.SIM_DCD_OUT, fh)
writeline(" fh.write('neigh_modify delay 1\n')", fh)
writeline(" fh.write('\n')", fh)
writeline(" fh.write('# -- Langevin section --\n')", fh)
writeline(" fh.write('# To use Langevin dynamics in LAMMPS you need both fix langevin and fix nve.\n')", fh)
writeline(" fh.write('# (See http://lammps.sandia.gov/doc/fix_langevin.html for details.)\n')", fh)
writeline(" fh.write('\n')", fh)
writeline(" fh.write('%s\n'%dampline)", fh)
writeline(" fh.write('fix fxnve all nve\n')", fh)
writeline(" fh.write('\n')", fh)
writeline(" ", fh)
writeline(" fh.write('thermo_style custom step temp pe etotal press vol epair ebond eangle edihed\n')", fh)
writeline(" fh.write('thermo %i # time interval for printing out thermo data\n')"%self.SIM_THERMO_OUT, fh)
writeline(" ", fh)
writeline(" fh.write('restart %i restart_nvt\n')"%self.SIM_NSTEPS, fh)
writeline(" fh.write('run %i\n')"%self.SIM_NSTEPS, fh)
writeline(" fh.write('write_restart system_after_nvt.rst\n')", fh)
writeline(" fh.write('\n')", fh)
writeline(" fh.write('\n')", fh)
def build_GaussianProcess_simulation_code(self):
"""
Creates a file which runs the actual LAMMPS simulation
"""
with open('run_simulation.py', 'w') as fh:
writeline('from shutil import copy', fh)
writeline('import os', fh)
writeline('import subprocess', fh)
writeline('#<><><><><><><><><><><><><><><><><><><><><><><><><><><><><>', fh)
writeline('# ', fh)
writeline('# ', fh)
writeline('# THIS CODE WAS DYNAMICALLY GENERATED BY CAMeLOT ', fh)
writeline('# Code generated at ' + str(time.strftime("%c")), fh)
writeline('# Generated by CAMeLOT version ' + self.MolTemplateParameters['CAMELOT_VERSION'] + "", fh)
writeline('#', fh)
writeline('#', fh)
writeline('# run_simulation.py', fh)
writeline('#', fh)
writeline('# This code should be called by the MATLAB code cg_opt_func.m', fh)
writeline('# and does the following things', fh)
writeline('#', fh)
writeline('# 1) Creates a new run.nvt.in configuration file (new SEED)', fh)
writeline('#', fh)
writeline('# 2) Moves that run.nvt.in file into the next iteration directory, which', fh)
writeline('# should already exist having been created by update_moltemplate.py, which', fh)
writeline('# itself should have been called immedietly preceding this function', fh)
writeline('#', fh)
writeline('# 3) Copies the standard run.in.min into the next iteration directory', fh)
writeline('#', fh)
writeline('# 4) Copies the standard system.lt file into the next iteration directory', fh)
writeline('#', fh)
writeline('# 5) Runs moltemplate on the system', fh)
writeline('#', fh)
writeline('# 6) launch the LAMMPS simulation in that directory', fh)
writeline('#', fh)
writeline('# 7) blocks until completed, then returns', fh)
writeline('#', fh)
writeline('#<><><><><><><><><><><><><><><><><><><><><><><><><><><><><>', fh)
writeline(' ', fh)
writeline("# Before we do that we have to figure out what iteration we're on", fh)
writeline("", fh)
writeline("try:", fh)
writeline(" with open('number_of_iterations.txt','r') as rfh:", fh)
writeline(" iteration = int(rfh.read())", fh)
writeline("except IOError:", fh)
writeline(" iteration = 0", fh)
writeline("", fh)
writeline('# 1) make run.nvt.in', fh)
writeline('os.system("%s generate_nvt.py")'%self.PYTHON_BIN, fh)
writeline('# 2) move run.nvt.in', fh)
writeline("os.rename('run.in.nvt', '%i/run.in.nvt'%iteration)", fh)
writeline('# 3) copy run.min.in', fh)
writeline("copy('run.in.min', '%i/run.in.min'%iteration)", fh)
writeline('# 4) copy system.lt', fh)
writeline("copy('system.lt', '%i/system.lt'%iteration)", fh)
writeline('# 5) run moltemplate', fh)
writeline('CWD = os.getcwd()', fh)
writeline('simdir = "%s/%i"%(CWD, iteration)',fh)
writeline("p = subprocess.Popen(['sh', '%s', 'system.lt'], cwd=simdir)"%(configs.MOLTEMPLATE3_LOCATION), fh)
writeline("p.wait()",fh)
writeline("p = subprocess.Popen(['%s', '-i', 'run.in.min'], cwd=simdir)"%(configs.LAMMPS_LOCATION), fh)
writeline("p.wait()",fh)
writeline("p = subprocess.Popen(['%s', '-i', 'run.in.nvt'], cwd=simdir)"%(configs.LAMMPS_LOCATION), fh)
writeline("p.wait()",fh)
def build_system_setup_file(self):
with open("system.lt", 'w') as fh:
writeline('# ', fh)
writeline('# ', fh)
writeline('# THIS CODE WAS DYNAMICALLY GENERATED BY CAMeLOT ', fh)
writeline('# Code generated at ' + str(time.strftime("%c")), fh)
writeline('# Generated by CAMeLOT version ' + self.MolTemplateParameters['CAMELOT_VERSION'], fh)
writeline('# ', fh)
writeline('# ', fh)
writeline('# system.lt', fh)
writeline('#<><><><><><><><><><><><><><><><><><><><><><><><><><><><><>', fh)
writeline(' ', fh)
writeline('import "stMol.lt"', fh)
writeline('write_once("Data Boundary") {', fh)
writeline(' 0 255.1 xlo xhi', fh)
writeline(' 0 255.1 ylo yhi', fh)
writeline(' 0 255.1 zlo zhi', fh)
writeline('}', fh)
writeline(' ',fh)
writeline('# Create a single peptide', fh)
writeline('peptide = new stMol [1].move(50.0, 100.0, 100.0) ', fh)
writeline(' ', fh)
writeline('#------------- END OF SYSTEM DEFINITION --------------', fh)
def build_nvt_minimization_file(self):
with open("run.in.min", 'w') as fh:
writeline('#<><><><><><><><><><><><><><><><><><><><><><><><><><><><><>', fh)
writeline('# ', fh)
writeline('# ', fh)
writeline('# THIS CODE WAS DYNAMICALLY GENERATED BY CAMeLOT ', fh)
writeline('# Code generated at ' + str(time.strftime("%c")), fh)
writeline('# Generated by CAMeLOT version ' + self.MolTemplateParameters['CAMELOT_VERSION'], fh)
writeline('# ', fh)
writeline('# ', fh)
writeline('# run.in.min', fh)
writeline('#<><><><><><><><><><><><><><><><><><><><><><><><><><><><><>', fh)
writeline('# -------- CAMeLOT MINIMIZATION: ---------', fh)
writeline('# This file defines a CAMeLOT minimization file', fh)
writeline(' ', fh)
writeline('# -- Init section --', fh)
writeline('include system.in.init', fh)
writeline(' ', fh)
writeline('# -- Atom definition section -- ', fh)
writeline('read_data system.data', fh)
writeline(' ', fh)
writeline('# -- Settings Section --', fh)
writeline('include system.in.settings', fh)
writeline(' ', fh)
writeline('# -- Run section --', fh)
writeline('dump 1 all custom 50 traj_min.lammpstrj id mol type x y z ix iy iz', fh)
writeline('minimize 1.0e-5 1.0e-7 10000 10000', fh)
writeline('write_restart system_after_min.rst',fh)
writeline(' ', fh)
writeline('#------------- END OF MINIMIZATION FILE --------------', fh)
def build_GaussianProcess_update_code(self):
"""
Creates a file called update_moltemplate.py which is a customized script which updates the
moltemplate file we created previously with the current interation of guesses
"""
(num_params,param_list) = self.determine_number_of_parameters_to_optimize()
with open('update_moltemplate.py','w') as fh:
writeline('import os', fh)
writeline('#<><><><><><><><><><><><><><><><><><><><><><><><><><><><><>', fh)
writeline('# ', fh)
writeline('# ', fh)
writeline('# THIS CODE WAS DYNAMICALLY GENERATED BY CAMeLOT ', fh)
writeline('# Code generated at ' + str(time.strftime("%c")) + "", fh)
writeline('# Generated by CAMeLOT version ' + self.MolTemplateParameters['CAMELOT_VERSION'] + "", fh)
writeline('# ', fh)
writeline('# ', fh)
writeline('# update_moltemplate.py', fh)
writeline('# \n', fh)
writeline('#<><><><><><><><><><><><><><><><><><><><><><><><><><><><><>', fh)
# for each parameter
fh.write("\n")
fh.write("try:\n")
fh.write(" with open('number_of_iterations.txt','r') as rfh:\n")
fh.write(" iteration = int(rfh.read())\n")
fh.write("except IOError:\n")
fh.write(" iteration = 0\n")
fh.write("\n")
# create a list of the parameters which are being optimized
fh.write('parameters = [');
for parameter_index in xrange(0,num_params):
if parameter_index+1 == num_params:
fh.write("'%s'] "%param_list[parameter_index])
else:
fh.write("'%s', "%param_list[parameter_index])
fh.write('\n')
fh.write('\n')
fh.write('current_guess = {}\n')
fh.write('\n')
fh.write('# for each parameter read in the current best guess\n')
fh.write('for i in parameters:\n')
fh.write(' try:\n')
fh.write(" with open(i+'_guess.txt','r') as rfh:\n")
fh.write(' current_guess[i] = float(rfh.read())\n')
fh.write(' except IOError:\n')
fh.write(" current_guess[i] = 1.01\n")
fh.write("\n")
fh.write(" # sigma parmeters have a coefficient...\n")
fh.write(" if i[0:3] == 'sig':\n")
fh.write(" try:\n")
fh.write(" with open('coff_' + i + '_guess.txt','r') as rfh:\n")
fh.write(" current_guess['coff_'+i] = float(rfh.read())\n")
fh.write(' except IOError:\n')
fh.write(" current_guess['coff_'+i] = 1.01\n")
fh.write("# The __TEMPLATE__ value updates a header iteration counter\n")
fh.write("current_guess['__TEMPLATE__'] = str(iteration)\n")
fh.write("\n")
fh.write("\n")
fh.write("# create next iteration directory\n")
fh.write("print 'Creating framework for iteration %i' % iteration\n")
fh.write("os.makedirs(str(iteration))\n")
fh.write("with open('" + self.MolTemplateParameters['OPT_MOLTEMPLATE_FILE'] +"','r') as fh:\n")
fh.write(" content = fh.readlines()\n")
fh.write("\n")
fh.write("newlines= []\n")
fh.write("for line in content:\n")
fh.write(" newline = line\n")
fh.write(" for key in current_guess:\n")
fh.write(" # note we add a leading space because we're looking for separate values\n")
fh.write(" newline = newline.replace(' ' + key, ' ' + str(current_guess[key]))\n")
fh.write(" newlines.append(newline)\n")
fh.write("\n")
fh.write("with open('%i/stMol.lt'%iteration,'w') as fh:\n")
fh.write(" for line in newlines:\n")
fh.write(" fh.write(line)\n")
fh.write("\n")
fh.write("\n")
def build_GaussianProcess_function_code(self):
"""
"""
with open('cg_opt_func.m','w') as fh:
writeline('function overlap = cg_opt_func(x, num_opt, type_opt)', fh)
writeline('%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%\n', fh)
writeline('% ', fh)
writeline('% ', fh)
writeline('% THIS CODE WAS DYNAMICALLY GENERATED BY CAMeLOT ', fh)
writeline('% Code generated at ' + str(time.strftime("%c")), fh)
writeline('% Generated by CAMeLOT version ' + self.MolTemplateParameters['CAMELOT_VERSION'], fh)
writeline('% ', fh)
writeline('% ', fh)
writeline('% cg_opt_func - Optimization function used by the BayesOpt to perform', fh)
writeline('% optimization', fh)
writeline('%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%', fh)
writeline(' ', fh)
writeline(' % Get current iteration', fh)
writeline(" iteration = load('number_of_iterations.txt');", fh)
writeline(" ", fh)
writeline(" % Add to iteration number", fh)
writeline(" iteration = iteration + 1;", fh)
writeline(" ", fh)
writeline(" % Print iteration number to file", fh)
writeline(" fileID=fopen('number_of_iterations.txt','w');", fh)
writeline(" fprintf(fileID, '%d', iteration);", fh)
writeline(" fclose(fileID);", fh)
writeline(" ", fh)
writeline(" % cycle through each parameter estimate and write to file", fh)
writeline(" for n = 1:num_opt", fh)
writeline(" tempname=sprintf('%s_guess.txt', type_opt{n});", fh)
writeline(" fileID=fopen(tempname, 'w');", fh)
writeline(" fprintf(fileID, '%4.2f\n', roundn(x(n),-2));", fh)
writeline(" fclose(fileID);", fh)
writeline(" %if we're dealing with a sigma value also write the coefficient", fh)
writeline(" ", fh)
writeline(" if strcmp(tempname(1:3), 'sig') == 1", fh)
writeline(" tempname=sprintf('coff_%s_guess.txt', type_opt{n});", fh)
writeline(" fileID=fopen(tempname, 'w');", fh)
writeline(" fprintf(fileID, '%4.2f\n', 2.5*roundn(x(n),-2));", fh)
writeline(" fclose(fileID);", fh)
writeline(" end", fh)
writeline(" end", fh)
writeline(" ", fh)
writeline(" ", fh)
# this line defines the creation and updating of the OPT_MOLTEMPLATE_FILE to include the new parameters we just
# wrote to file. update_moltemplate.py is a stand alone script which is
writeline(" % this line will run a custom generated script which will update ", fh)
writeline(" % the moltemplate file with the current guesses ", fh)
writeline(" system('" + self.PYTHON_BIN + " update_moltemplate.py');", fh)
writeline("", fh)
writeline(" % this line will launch the simulation and then poll until the ", fh)
writeline(" % simulation has finished ", fh)
writeline(" system('" + self.PYTHON_BIN + " run_simulation.py');", fh)
writeline("", fh)
writeline(" % this line will run custom generated script which will compares", fh)
writeline(" % the CG simulation against the all atom simulation ", fh)
writeline(" overlap = evaluate_simulation(iteration, %i)" % len(self.sequence_vector), fh)
writeline("", fh)
writeline(" % Save the current overlap score to a textfile for convenience ", fh)
writeline(" fid = fopen('goodness.txt', 'at');", fh)
writeline(" fprintf(fid, '%i %4.2f\n', iteration, overlap);", fh)
writeline(" fclose(fid);", fh)
writeline(' ', fh)
writeline('% END OF FILE --------------------------------', fh)
def build_GaussianProcess_runner_code(self):
"""
The Gaussiang process runner code (found in cg_opt.m) is the primary optimization code which is runs in MATLAB
and acts as the framework around which the Gaussian process optimization occurs.
This code itself calls another function which is custom generated by CAMeLOT (cg_opt_func.m)
"""
(num_params,param_list) = self.determine_number_of_parameters_to_optimize()
parameter_min_max = self.determine_parameter_min_and_max()
# this whole function generates a bunch of runner code - specifically it creates
# 1) A matlab parent script
# 2) A matlab function called by the parent script to execute the Gaussian process
with open('cg_opt.m','w') as fh:
fh.write('%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%\n')
fh.write('% \n')
fh.write('% \n')
fh.write('% THIS CODE WAS DYNAMICALLY GENERATED BY CAMeLOT \n')
fh.write('% Code generated at ' + str(time.strftime("%c")) + "\n")
fh.write('% Generated by CAMeLOT version ' + self.MolTemplateParameters['CAMELOT_VERSION'] + "\n")
fh.write('% \n')
fh.write('% cg_opt - Optimization framework script which runs the BayesOpt package\n')
fh.write('% to perform Gaussian Process Bayesian optimization\n')
fh.write('%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%\n')
fh.write('\n')
fh.write('\n')
fh.write('% Initialize Gaussian process\n')
fh.write("addpath %s\n" % self.GP_MATLAB_PATH)
fh.write("startup % this initializes the GP framework\n")
fh.write('\n')
fh.write('delete goodness.txt % delete the goodness file if it exists\n')
fh.write("num_opt = %i;\n" % num_params)
fh.write('% Set parameters to be optimized\n')
# dynamically construct a matlab cell with the parameter types to be optimized (will match the names
# used throughout the process)
fh.write("type_opt = {'")
for parameter_index in xrange(0,num_params):
# if we're on the last parameter
if parameter_index+1 == num_params:
fh.write("%s'" % param_list[parameter_index])
else:
fh.write("%s', '" % param_list[parameter_index])
fh.write("};\n")
# write initial parameters
writeline('for n = 1:num_opt',fh)
writeline(" tempname=sprintf('%s_guess.txt', type_opt{n});", fh)
writeline(" fileID=fopen(tempname, 'w');", fh)
writeline(" fprintf(fileID, '%4.2f\n');"%0.0, fh)
writeline(" % if we're dealing with a sigma value also write the coefficient", fh)
writeline(" ", fh)
writeline(" if strcmp(tempname(1:3), 'sig') == 1", fh)
writeline(" tempname=sprintf('coff_%s_guess.txt', type_opt{n});", fh)
writeline(" fileID=fopen(tempname, 'w');", fh)
writeline(" fprintf(fileID, '%4.2f\n');" % 0.0 , fh)
writeline(" fclose(fileID);", fh)
writeline(" end", fh)
writeline("end", fh)
writeline("fileID=fopen('number_of_iterations.txt','w');", fh)
writeline("fprintf(fileID, '%i');" % 0, fh)
writeline("fclose(fileID);", fh)
# write the actual anonymous function call which is to be used. Note that we're essentially using a closure where
# we're defining num_opt and type_opt HERE as we define the function, but when F is called it only
# accepts 'x' (vector of values to be optimized) while num_opt and type_opt are already defined
fh.write('F = @(x)cg_opt_func(x, num_opt, type_opt);\n');
# write the options
fh.write('\n')
fh.write('opt = bo_default_opt();\n')
fh.write('opt.dims = num_opt; \n')
fh.write('\n')
# dynamically construct matlab code defining the minimum and maximum values for each
# of the parameter
fh.write('% Set min and max for optimization procedure\n')
for parameter_index in xrange(0,num_params):
parameter_name = param_list[parameter_index]
fh.write('opt.mins(%i) = %3.3f;\n' % (parameter_index+1,parameter_min_max[parameter_name][0]))
fh.write('opt.maxes(%i) = %3.3f;\n' % (parameter_index+1,parameter_min_max[parameter_name][1]))
fh.write('\n')
# set the number of iterations
fh.write('% Set number of optimization iterations [OPT_ITERATIONS]\n')
fh.write('opt.max_iters = %i;\n'%(self.KeyFileObj.OPT_ITERATIONS))
fh.write('opt.save_trace = true;\n')
fh.write('\n')
# write the actual optimization call out!
fh.write('% run optimization\n')
fh.write("[ms,mv,T] = bayesopt(F,opt);\n")
fh.write('\n')
fh.write("fprintf('\\n**Optimization finished.\\n');\n")
fh.write("fprintf('Minimum function value found in grid was \%f\\n',mv);\n")
fh.write("fprintf('Point with minimum value was...\\n');\n")
fh.write("disp(ms);\n")