1 | import numpy
|
---|
2 | import copy
|
---|
3 | from fielddisplay import fielddisplay
|
---|
4 | from EnumDefinitions import *
|
---|
5 | from StringToEnum import StringToEnum
|
---|
6 | from checkfield import *
|
---|
7 | from WriteData import *
|
---|
8 |
|
---|
9 | class inversion(object):
|
---|
10 | """
|
---|
11 | INVERSION class definition
|
---|
12 |
|
---|
13 | Usage:
|
---|
14 | inversion=inversion();
|
---|
15 | """
|
---|
16 |
|
---|
17 | def __init__(self): # {{{
|
---|
18 | self.iscontrol = 0
|
---|
19 | self.tao = 0
|
---|
20 | self.incomplete_adjoint = 0
|
---|
21 | self.control_parameters = float('NaN')
|
---|
22 | self.nsteps = 0
|
---|
23 | self.maxiter_per_step = float('NaN')
|
---|
24 | self.cost_functions = float('NaN')
|
---|
25 | self.cost_functions_coefficients = float('NaN')
|
---|
26 | self.gradient_scaling = float('NaN')
|
---|
27 | self.cost_function_threshold = 0
|
---|
28 | self.min_parameters = float('NaN')
|
---|
29 | self.max_parameters = float('NaN')
|
---|
30 | self.step_threshold = float('NaN')
|
---|
31 | self.vx_obs = float('NaN')
|
---|
32 | self.vy_obs = float('NaN')
|
---|
33 | self.vz_obs = float('NaN')
|
---|
34 | self.vel_obs = float('NaN')
|
---|
35 | self.thickness_obs = float('NaN')
|
---|
36 |
|
---|
37 | #set defaults
|
---|
38 | self.setdefaultparameters()
|
---|
39 |
|
---|
40 | #}}}
|
---|
41 | def __repr__(self): # {{{
|
---|
42 | string=' inversion parameters:'
|
---|
43 | string="%s\n%s"%(string,fielddisplay(self,'iscontrol','is inversion activated?'))
|
---|
44 | string="%s\n%s"%(string,fielddisplay(self,'incomplete_adjoint','1: linear viscosity, 0: non-linear viscosity'))
|
---|
45 | string="%s\n%s"%(string,fielddisplay(self,'control_parameters','ex: {''FrictionCoefficient''}, or {''MaterialsRheologyBbar''}'))
|
---|
46 | string="%s\n%s"%(string,fielddisplay(self,'nsteps','number of optimization searches'))
|
---|
47 | string="%s\n%s"%(string,fielddisplay(self,'cost_functions','indicate the type of response for each optimization step'))
|
---|
48 | string="%s\n%s"%(string,fielddisplay(self,'cost_functions_coefficients','cost_functions_coefficients applied to the misfit of each vertex and for each control_parameter'))
|
---|
49 | string="%s\n%s"%(string,fielddisplay(self,'cost_function_threshold','misfit convergence criterion. Default is 1%, NaN if not applied'))
|
---|
50 | string="%s\n%s"%(string,fielddisplay(self,'maxiter_per_step','maximum iterations during each optimization step'))
|
---|
51 | string="%s\n%s"%(string,fielddisplay(self,'gradient_scaling','scaling factor on gradient direction during optimization, for each optimization step'))
|
---|
52 | string="%s\n%s"%(string,fielddisplay(self,'step_threshold','decrease threshold for misfit, default is 30%'))
|
---|
53 | string="%s\n%s"%(string,fielddisplay(self,'min_parameters','absolute minimum acceptable value of the inversed parameter on each vertex'))
|
---|
54 | string="%s\n%s"%(string,fielddisplay(self,'max_parameters','absolute maximum acceptable value of the inversed parameter on each vertex'))
|
---|
55 | string="%s\n%s"%(string,fielddisplay(self,'vx_obs','observed velocity x component [m/yr]'))
|
---|
56 | string="%s\n%s"%(string,fielddisplay(self,'vy_obs','observed velocity y component [m/yr]'))
|
---|
57 | string="%s\n%s"%(string,fielddisplay(self,'vel_obs','observed velocity magnitude [m/yr]'))
|
---|
58 | string="%s\n%s"%(string,fielddisplay(self,'thickness_obs','observed thickness [m]'))
|
---|
59 | string="%s\n%s"%(string,'Available cost functions:')
|
---|
60 | string="%s\n%s"%(string,' 101: SurfaceAbsVelMisfit')
|
---|
61 | string="%s\n%s"%(string,' 102: SurfaceRelVelMisfit')
|
---|
62 | string="%s\n%s"%(string,' 103: SurfaceLogVelMisfit')
|
---|
63 | string="%s\n%s"%(string,' 104: SurfaceLogVxVyMisfit')
|
---|
64 | string="%s\n%s"%(string,' 105: SurfaceAverageVelMisfit')
|
---|
65 | string="%s\n%s"%(string,' 201: ThicknessAbsMisfit')
|
---|
66 | string="%s\n%s"%(string,' 501: DragCoefficientAbsGradient')
|
---|
67 | string="%s\n%s"%(string,' 502: RheologyBbarAbsGradient')
|
---|
68 | string="%s\n%s"%(string,' 503: ThicknessAbsGradient')
|
---|
69 | return string
|
---|
70 | #}}}
|
---|
71 | def setdefaultparameters(self): # {{{
|
---|
72 |
|
---|
73 | #default is incomplete adjoint for now
|
---|
74 | self.incomplete_adjoint=1
|
---|
75 |
|
---|
76 | #parameter to be inferred by control methods (only
|
---|
77 | #drag and B are supported yet)
|
---|
78 | self.control_parameters='FrictionCoefficient'
|
---|
79 |
|
---|
80 | #number of steps in the control methods
|
---|
81 | self.nsteps=20
|
---|
82 |
|
---|
83 | #maximum number of iteration in the optimization algorithm for
|
---|
84 | #each step
|
---|
85 | self.maxiter_per_step=20*numpy.ones(self.nsteps)
|
---|
86 |
|
---|
87 | #the inversed parameter is updated as follows:
|
---|
88 | #new_par=old_par + gradient_scaling(n)*C*gradient with C in [0 1];
|
---|
89 | #usually the gradient_scaling must be of the order of magnitude of the
|
---|
90 | #inversed parameter (10^8 for B, 50 for drag) and can be decreased
|
---|
91 | #after the first iterations
|
---|
92 | self.gradient_scaling=50*numpy.ones((self.nsteps,1))
|
---|
93 |
|
---|
94 | #several responses can be used:
|
---|
95 | self.cost_functions=101
|
---|
96 |
|
---|
97 | #step_threshold is used to speed up control method. When
|
---|
98 | #misfit(1)/misfit(0) < self.step_threshold, we go directly to
|
---|
99 | #the next step
|
---|
100 | self.step_threshold=.7*numpy.ones(self.nsteps) #30 per cent decrement
|
---|
101 |
|
---|
102 | #cost_function_threshold is a criteria to stop the control methods.
|
---|
103 | #if J[n]-J[n-1]/J[n] < criteria, the control run stops
|
---|
104 | #NaN if not applied
|
---|
105 | self.cost_function_threshold=float('NaN') #not activated
|
---|
106 |
|
---|
107 | return self
|
---|
108 | #}}}
|
---|
109 | def checkconsistency(self,md,solution,analyses): # {{{
|
---|
110 |
|
---|
111 | #Early return
|
---|
112 | if not self.iscontrol:
|
---|
113 | return md
|
---|
114 |
|
---|
115 | num_controls=numpy.size(md.inversion.control_parameters)
|
---|
116 | num_costfunc=numpy.size(md.inversion.cost_functions)
|
---|
117 |
|
---|
118 | md = checkfield(md,'fieldname','inversion.iscontrol','values',[0,1])
|
---|
119 | md = checkfield(md,'fieldname','inversion.tao','values',[0,1])
|
---|
120 | md = checkfield(md,'fieldname','inversion.incomplete_adjoint','values',[0,1])
|
---|
121 | md = checkfield(md,'fieldname','inversion.control_parameters','cell',1,'values',['BalancethicknessThickeningRate','FrictionCoefficient','MaterialsRheologyBbar','DamageDbar','Vx','Vy'])
|
---|
122 | md = checkfield(md,'fieldname','inversion.nsteps','numel',[1],'>=',0)
|
---|
123 | md = checkfield(md,'fieldname','inversion.maxiter_per_step','size',[md.inversion.nsteps],'>=',0)
|
---|
124 | md = checkfield(md,'fieldname','inversion.step_threshold','size',[md.inversion.nsteps])
|
---|
125 | md = checkfield(md,'fieldname','inversion.cost_functions','size',[num_costfunc],'values',[101,102,103,104,105,201,501,502,503,504,505])
|
---|
126 | md = checkfield(md,'fieldname','inversion.cost_functions_coefficients','size',[md.mesh.numberofvertices,num_costfunc],'>=',0)
|
---|
127 | md = checkfield(md,'fieldname','inversion.gradient_scaling','size',[md.inversion.nsteps,num_controls])
|
---|
128 | md = checkfield(md,'fieldname','inversion.min_parameters','size',[md.mesh.numberofvertices,num_controls])
|
---|
129 | md = checkfield(md,'fieldname','inversion.max_parameters','size',[md.mesh.numberofvertices,num_controls])
|
---|
130 |
|
---|
131 | #Only SSA, HO and FS are supported right now
|
---|
132 | if solution==StressbalanceSolutionEnum():
|
---|
133 | if not (md.flowequation.isSSA or md.flowequation.isHO or md.flowequation.isFS):
|
---|
134 | md.checkmessage("'inversion can only be performed for SSA, HO or FS ice flow models");
|
---|
135 |
|
---|
136 | if solution==BalancethicknessSolutionEnum():
|
---|
137 | md = checkfield(md,'fieldname','inversion.thickness_obs','size',[md.mesh.numberofvertices],'NaN',1)
|
---|
138 | else:
|
---|
139 | md = checkfield(md,'fieldname','inversion.vx_obs','size',[md.mesh.numberofvertices],'NaN',1)
|
---|
140 | md = checkfield(md,'fieldname','inversion.vy_obs','size',[md.mesh.numberofvertices],'NaN',1)
|
---|
141 |
|
---|
142 | return md
|
---|
143 | # }}}
|
---|
144 | def marshall(self,md,fid): # {{{
|
---|
145 |
|
---|
146 | yts=365.0*24.0*3600.0
|
---|
147 |
|
---|
148 | WriteData(fid,'object',self,'fieldname','iscontrol','format','Boolean')
|
---|
149 | WriteData(fid,'object',self,'fieldname','tao','format','Boolean')
|
---|
150 | WriteData(fid,'object',self,'fieldname','incomplete_adjoint','format','Boolean')
|
---|
151 | if not self.iscontrol:
|
---|
152 | return
|
---|
153 | WriteData(fid,'object',self,'fieldname','nsteps','format','Integer')
|
---|
154 | WriteData(fid,'object',self,'fieldname','maxiter_per_step','format','DoubleMat','mattype',3)
|
---|
155 | WriteData(fid,'object',self,'fieldname','cost_functions_coefficients','format','DoubleMat','mattype',1)
|
---|
156 | WriteData(fid,'object',self,'fieldname','gradient_scaling','format','DoubleMat','mattype',3)
|
---|
157 | WriteData(fid,'object',self,'fieldname','cost_function_threshold','format','Double')
|
---|
158 | WriteData(fid,'object',self,'fieldname','min_parameters','format','DoubleMat','mattype',3)
|
---|
159 | WriteData(fid,'object',self,'fieldname','max_parameters','format','DoubleMat','mattype',3)
|
---|
160 | WriteData(fid,'object',self,'fieldname','step_threshold','format','DoubleMat','mattype',3)
|
---|
161 | WriteData(fid,'object',self,'fieldname','vx_obs','format','DoubleMat','mattype',1,'scale',1./yts)
|
---|
162 | WriteData(fid,'object',self,'fieldname','vy_obs','format','DoubleMat','mattype',1,'scale',1./yts)
|
---|
163 | WriteData(fid,'object',self,'fieldname','vz_obs','format','DoubleMat','mattype',1,'scale',1./yts)
|
---|
164 | WriteData(fid,'object',self,'fieldname','thickness_obs','format','DoubleMat','mattype',1)
|
---|
165 |
|
---|
166 | #process control parameters
|
---|
167 | num_control_parameters=len(self.control_parameters)
|
---|
168 | data=numpy.array([StringToEnum(control_parameter)[0] for control_parameter in self.control_parameters]).reshape(1,-1)
|
---|
169 | WriteData(fid,'data',data,'enum',InversionControlParametersEnum(),'format','DoubleMat','mattype',3)
|
---|
170 | WriteData(fid,'data',num_control_parameters,'enum',InversionNumControlParametersEnum(),'format','Integer')
|
---|
171 |
|
---|
172 | #process cost functions
|
---|
173 | num_cost_functions=numpy.size(self.cost_functions)
|
---|
174 | data=copy.deepcopy(self.cost_functions)
|
---|
175 | data=[SurfaceAbsVelMisfitEnum() if x==101 else x for x in data]
|
---|
176 | data=[SurfaceRelVelMisfitEnum() if x==102 else x for x in data]
|
---|
177 | data=[SurfaceLogVelMisfitEnum() if x==103 else x for x in data]
|
---|
178 | data=[SurfaceLogVxVyMisfitEnum() if x==104 else x for x in data]
|
---|
179 | data=[SurfaceAverageVelMisfitEnum() if x==105 else x for x in data]
|
---|
180 | data=[ThicknessAbsMisfitEnum() if x==201 else x for x in data]
|
---|
181 | data=[DragCoefficientAbsGradientEnum() if x==501 else x for x in data]
|
---|
182 | data=[RheologyBbarAbsGradientEnum() if x==502 else x for x in data]
|
---|
183 | data=[ThicknessAbsGradientEnum() if x==503 else x for x in data]
|
---|
184 | data=[ThicknessAlongGradientEnum() if x==504 else x for x in data]
|
---|
185 | data=[ThicknessAcrossGradientEnum() if x==505 else x for x in data]
|
---|
186 | WriteData(fid,'data',numpy.array(data).reshape(1,-1),'enum',InversionCostFunctionsEnum(),'format','DoubleMat','mattype',3)
|
---|
187 | WriteData(fid,'data',num_cost_functions,'enum',InversionNumCostFunctionsEnum(),'format','Integer')
|
---|
188 | # }}}
|
---|