code.rst 22.1 KB
Newer Older
1

Martin Reinecke's avatar
Martin Reinecke committed
2
=============
3 4
Code Overview
=============
5

Martin Reinecke's avatar
Martin Reinecke committed
6 7 8 9

Executive summary
=================

10 11
The fundamental building blocks required for IFT computations are best
recognized from a large distance, ignoring all technical details.
12

13 14
From such a perspective,

Martin Reinecke's avatar
tweaks  
Martin Reinecke committed
15
- IFT problems largely consist of the combination of several high dimensional
16
  *minimization* problems.
Philipp Arras's avatar
Philipp Arras committed
17
- Within NIFTy, *operators* are used to define the characteristic equations and
Martin Reinecke's avatar
tweaks  
Martin Reinecke committed
18
  properties of the problems.
Martin Reinecke's avatar
Martin Reinecke committed
19 20
- The equations are built mostly from the application of *linear operators*,
  but there may also be nonlinear functions involved.
21
- The unknowns in the equations represent either continuous physical *fields*,
Martin Reinecke's avatar
Martin Reinecke committed
22 23 24 25
  or they are simply individual measured *data points*.
- Discretized *fields* have geometrical information (like locations and volume
  elements) associated with every entry; this information is called the field's
  *domain*.
26 27 28 29 30 31 32 33 34 35 36 37 38

In the following sections, the concepts briefly presented here will be
discussed in more detail; this is done in reversed order of their introduction,
to avoid forward references.


Domains
=======


Abstract base class
-------------------

39 40
.. currentmodule:: nifty5.domains.domain

Philipp Arras's avatar
Philipp Arras committed
41
One of the fundamental building blocks of the NIFTy5 framework is the *domain*.
42
Its required capabilities are expressed by the abstract :py:class:`Domain` class.
43 44 45 46 47 48 49 50 51 52 53 54
A domain must be able to answer the following queries:

- its total number of data entries (pixels), which is accessible via the
  :attr:`~Domain.size` property
- the shape of the array that is supposed to hold these data entries
  (obtainable by means of the :attr:`~Domain.shape` property)
- equality comparison to another :class:`Domain` instance


Unstructured domains
--------------------

55 56
.. currentmodule:: nifty5.domains.unstructured_domain

57 58 59 60 61 62 63 64 65 66 67
Domains can be either *structured* (i.e. there is geometrical information
associated with them, like position in space and volume factors),
or *unstructured* (meaning that the data points have no associated manifold).

Unstructured domains can be described by instances of NIFTy's
:class:`UnstructuredDomain` class.


Structured domains
------------------

68 69
.. currentmodule:: nifty5.domains.structured_domain

70 71 72 73 74
In contrast to unstructured domains, these domains have an assigned geometry.
NIFTy requires them to provide the volume elements of their grid cells.
The additional methods are specified in the abstract class
:class:`StructuredDomain`:

Martin Reinecke's avatar
Martin Reinecke committed
75
- The properties :attr:`~StructuredDomain.scalar_dvol`,
76 77 78 79
  :attr:`~StructuredDomain.dvol`, and  :attr:`~StructuredDomain.total_volume`
  provide information about the domain's pixel volume(s) and its total volume.
- The property :attr:`~StructuredDomain.harmonic` specifies whether a domain
  is harmonic (i.e. describes a frequency space) or not
80
- If the domain is harmonic, the methods
81 82 83 84 85 86 87 88
  :meth:`~StructuredDomain.get_k_length_array`,
  :meth:`~StructuredDomain.get_unique_k_lengths`, and
  :meth:`~StructuredDomain.get_fft_smoothing_kernel_function` provide absolute
  distances of the individual grid cells from the origin and assist with
  Gaussian convolution.

NIFTy comes with several concrete subclasses of :class:`StructuredDomain`:

89 90
.. currentmodule:: nifty5.domains

Philipp Arras's avatar
Philipp Arras committed
91
- :class:`~rg_space.RGSpace` represents a regular Cartesian grid with an arbitrary
92
  number of dimensions, which is supposed to be periodic in each dimension.
93 94
- :class:`~log_rg_space.LogRGSpace` implements a Cartesian grid wit logarithimcally
  spaced bins and an arbitrary number of dimensions.
Philipp Arras's avatar
Philipp Arras committed
95 96
- :class:`~hp_space.HPSpace` and :class:`~gl_space.GLSpace` describe pixelisations of the
  2-sphere; their counterpart in harmonic space is :class:`~lm_space.LMSpace`, which
97
  contains spherical harmonic coefficients.
Philipp Arras's avatar
Philipp Arras committed
98
- :class:`~power_space.PowerSpace` is used to describe one-dimensional power spectra.
99

100 101 102
Among these, :class:`~rg_space.RGSpace` and :class:`~log_rg_space.LogRGSpace` can
be harmonic or not (depending on constructor arguments),
:class:`~gl_space.GLSpace`, :class:`~hp_space.HPSpace`,
Philipp Arras's avatar
Philipp Arras committed
103 104
and :class:`~power_space.PowerSpace` are pure position domains (i.e.
nonharmonic), and :class:`~lm_space.LMSpace` is always harmonic.
105 106 107 108 109 110


Combinations of domains
=======================

The fundamental classes described above are often sufficient to specify the
111 112
domain of a field. In some cases, however, it will be necessary to define the
field on a product of elementary domains instead of a single one.
Philipp Arras's avatar
Philipp Arras committed
113
More sophisticated operators also require a set of several such fields.
114 115
Some examples are:

Philipp Arras's avatar
Philipp Arras committed
116 117 118
- sky emission depending on location and energy. This could be represented by a
  product of an :class:`~hp_space.HPSpace` (for location) with an
  :class:`~rg_space.RGSpace` (for energy).
119
- a polarized field, which could be modelled as a product of any structured
Martin Reinecke's avatar
Martin Reinecke committed
120
  domain (representing location) with a four-element
Philipp Arras's avatar
Philipp Arras committed
121
  :class:`~unstructured_domain.UnstructuredDomain` holding Stokes I, Q, U and V components.
122
- a model for the sky emission, which holds both the current realisation
Martin Reinecke's avatar
Martin Reinecke committed
123 124
  (on a harmonic domain) and a few inferred model parameters (e.g. on an
  unstructured grid).
125

Philipp Arras's avatar
Philipp Arras committed
126
.. currentmodule:: nifty5
127

128 129 130 131 132
Consequently, NIFTy defined a class called :class:`~domain_tuple.DomainTuple`
holding a sequence of :class:`~domains.domain.Domain` objects. The full domain is
specified as the product of all elementary domains. Thus, an instance of
:class:`~domain_tuple.DomainTuple` would be suitable to describe the former two
examples above. In principle, a :class:`~domain_tuple.DomainTuple`
Philipp Arras's avatar
Philipp Arras committed
133
can even be empty, which implies that the field living on it is a scalar.
Martin Reinecke's avatar
Martin Reinecke committed
134

135 136 137 138 139
.. Consequently, NIFTy defines a class called :class:`~domain_tuple.DomainTuple`
.. holding a sequence of :class:`~domains.domain.Domain` objects, which is used to
.. specify full field domains. In principle, a :class:`~domain_tuple.DomainTuple`
.. can even be empty, which implies that the field living on it is a scalar.

Philipp Arras's avatar
Philipp Arras committed
140
A :class:`~domain_tuple.DomainTuple` supports iteration and indexing, and also
141
provides the properties :attr:`~domain_tuple.DomainTuple.shape` and
Philipp Arras's avatar
Philipp Arras committed
142 143 144 145 146
:attr:`~domain_tuple.DomainTuple.size` in analogy to the elementary
:class:`~domains.domain.Domain`.

An aggregation of several :class:`~domain_tuple.DomainTuple` s, each member
identified by a name, is described by the :class:`~multi_domain.MultiDomain`
147 148 149 150
class. In contrast to a :class:`~domain_tuple.DomainTuple` a
:class:`~multi_domain.MultiDomain` is a collection and does not define the
product space of its elements.  It would be the adequate space to use in the
latter of above's examples.
151 152 153 154

Fields
======

Martin Reinecke's avatar
Martin Reinecke committed
155 156 157
Fields on a single DomainTuple
------------------------------

Philipp Arras's avatar
Philipp Arras committed
158
A :class:`~field.Field` object consists of the following components:
159

Philipp Arras's avatar
Philipp Arras committed
160
- a domain in form of a :class:`~domain_tuple.DomainTuple` object
161 162 163
- a data type (e.g. numpy.float64)
- an array containing the actual values

Martin Reinecke's avatar
Martin Reinecke committed
164
Usually, the array is stored in the form of a ``numpy.ndarray``, but for very
Martin Reinecke's avatar
Martin Reinecke committed
165 166 167
resource-intensive tasks NIFTy also provides an alternative storage method to
be used with distributed memory processing.

168 169 170 171 172 173 174 175 176 177 178
Fields support a wide range of arithmetic operations, either involving
two fields of equal domains or a field and a scalar. Arithmetic operations are
performed point-wise, and the returned field has the same domain as the input field(s). Available operators are addition ("+"), subtraction ("-"),
multiplication ("*"), division ("/"), floor division ("//") and
exponentiation ("**"). Inplace operators ("+=", etc.) are not supported.
Further, boolean operators, performing a point wise comparison of a field with
either another field of equal domain or a scalar, are available as well. These
include equals ("=="), not equals ("!="), less ("<"), less or equal ("<="),
greater (">") and greater or equal (">=). The domain of the field returned equals
that of the input field(s), while the stored data is of boolean type.

Martin Reinecke's avatar
Martin Reinecke committed
179 180
Contractions (like summation, integration, minimum/maximum, computation of
statistical moments) can be carried out either over an entire field (producing
Philipp Arras's avatar
Philipp Arras committed
181
a scalar result) or over sub-domains (resulting in a field defined on a smaller
182
domain). Scalar products of two fields can also be computed easily as well.
Martin Reinecke's avatar
Martin Reinecke committed
183
See the documentation of :class:`~field.Field` for details.
Martin Reinecke's avatar
Martin Reinecke committed
184 185 186

There is also a set of convenience functions to generate fields with constant
values or fields filled with random numbers according to a user-specified
187
distribution: :attr:`~sugar.full`, :attr:`~sugar.from_random`.
Martin Reinecke's avatar
Martin Reinecke committed
188

Martin Reinecke's avatar
Martin Reinecke committed
189 190 191 192 193
Like almost all NIFTy objects, fields are immutable: their value or any other
attribute cannot be modified after construction. To manipulate a field in ways
that are not covered by the provided standard operations, its data content must
be extracted first, then changed, and a new field has to be created from the
result.
Martin Reinecke's avatar
Martin Reinecke committed
194

195
Fields defined on a MultiDomain
Martin Reinecke's avatar
Martin Reinecke committed
196
-------------------------------
Martin Reinecke's avatar
Martin Reinecke committed
197

Philipp Arras's avatar
Philipp Arras committed
198 199 200
The :class:`~multi_field.MultiField` class can be seen as a dictionary of
individual :class:`~field.Field` s, each identified by a name, which is defined
on a :class:`~multi_domain.MultiDomain`.
Martin Reinecke's avatar
Martin Reinecke committed
201 202 203 204 205


Operators
=========

Philipp Arras's avatar
Philipp Arras committed
206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223
All transformations between different NIFTy fields are expressed in the form of
:class:`~operators.operator.Operator` objects. The interface of this class is
rather minimalistic: it has a property called
:attr:`~operators.operator.Operator.domain` which returns a
:class:`~domain_tuple.DomainTuple` or :class:`~multi_domain.MultiDomain` object
specifying the structure of the :class:`~field.Field` or
:class:`~multi_field.MultiField` it expects as input, another property
:attr:`~operators.operator.Operator.target` describing its output, and finally
an overloaded :attr:`~operators.operator.Operator.apply` method, which can take:

- a :class:`~field.Field`/:class:`~multi_field.MultiField` object, in which case
  it returns the transformed :class:`~field.Field`/:class:`~multi_field.MultiField`.
- a :class:`~linearization.Linearization` object, in which case it returns the
  transformed :class:`~linearization.Linearization`.

This is the interface that all objects derived from
:class:`~operators.operator.Operator` must implement. In addition,
:class:`~operators.operator.Operator` objects can be added/subtracted,
Philipp Arras's avatar
Docs pa  
Philipp Arras committed
224 225 226
multiplied, chained (via the :attr:`__call__` method or the `@` operator) and
support point-wise application of functions like :class:`exp()`, :class:`log()`,
:class:`sqrt()`, :class:`conjugate()`.
Martin Reinecke's avatar
Martin Reinecke committed
227

228

Martin Reinecke's avatar
Martin Reinecke committed
229 230 231
Advanced operators
------------------

Philipp Arras's avatar
Philipp Arras committed
232
NIFTy provides a library of commonly employed operators which can be used for
Martin Reinecke's avatar
Martin Reinecke committed
233 234
specific inference problems. Currently these are:

Philipp Arras's avatar
Docs pa  
Philipp Arras committed
235 236
.. currentmodule:: nifty5.library

237
- :class:`~smooth_linear_amplitude.SLAmplitude`, which returns a smooth power spectrum.
Philipp Arras's avatar
Docs pa  
Philipp Arras committed
238
- :class:`~inverse_gamma_operator.InverseGammaOperator`, which models point sources which are
Philipp Arras's avatar
Philipp Arras committed
239
  distributed according to a inverse-gamma distribution.
240 241 242 243
- :class:`~correlated_fields.CorrelatedField`, which models a diffuse field whose correlation
  structure is described by an amplitude operator.
.. - :class:`~correlated_fields.CorrelatedField`, which models a diffuse log-normal field. It takes an
   amplitude operator to specify the correlation structure of the field.
Martin Reinecke's avatar
Martin Reinecke committed
244 245


246 247 248
Linear Operators
================

Philipp Arras's avatar
Docs pa  
Philipp Arras committed
249 250 251 252 253 254 255 256
.. currentmodule:: nifty5.operators

A linear operator (represented by NIFTy5's abstract
:class:`~linear_operator.LinearOperator` class) is derived from
:class:`~operator.Operator` and can be interpreted as an (implicitly defined)
matrix. Since its operation is linear, it can provide some additional
functionality which is not available for the more generic
:class:`~operator.Operator` class.
257

Martin Reinecke's avatar
Martin Reinecke committed
258

Martin Reinecke's avatar
Martin Reinecke committed
259 260
Linear Operator basics
----------------------
Martin Reinecke's avatar
Martin Reinecke committed
261

Philipp Arras's avatar
Docs pa  
Philipp Arras committed
262
There are four basic ways of applying an operator :math:`A` to a field :math:`s`:
263

Philipp Arras's avatar
Docs pa  
Philipp Arras committed
264 265 266 267
- direct application: :math:`A(s)`
- adjoint application: :math:`A^\dagger (s)`
- inverse application: :math:`A^{-1} (s)`
- adjoint inverse application: :math:`(A^\dagger)^{-1} (s)`
268

Philipp Arras's avatar
Docs pa  
Philipp Arras committed
269 270
Note: The inverse of the adjoint of a linear map and the adjoint of the inverse
of a linear map (if all those exist) are the same.
271

Martin Reinecke's avatar
Martin Reinecke committed
272 273
These different actions of a linear operator ``Op`` on a field ``f`` can be
invoked in various ways:
274 275 276 277 278 279 280 281

- direct multiplication: ``Op(f)`` or ``Op.times(f)`` or ``Op.apply(f, Op.TIMES)``
- adjoint multiplication: ``Op.adjoint_times(f)`` or ``Op.apply(f, Op.ADJOINT_TIMES)``
- inverse multiplication: ``Op.inverse_times(f)`` or ``Op.apply(f, Op.INVERSE_TIMES)``
- adjoint inverse multiplication: ``Op.adjoint_inverse_times(f)`` or ``Op.apply(f, Op.ADJOINT_INVERSE_TIMES)``

Operator classes defined in NIFTy may implement an arbitrary subset of these
four operations. This subset can be queried using the
Philipp Arras's avatar
Docs pa  
Philipp Arras committed
282
:attr:`~linear_operator.LinearOperator.capability` property.
283 284

If needed, the set of supported operations can be enhanced by iterative
Philipp Arras's avatar
Docs pa  
Philipp Arras committed
285 286 287 288
inversion methods; for example, an operator defining direct and adjoint
multiplication could be enhanced by this approach to support the complete set.
This functionality is provided by NIFTy's
:class:`~inversion_enabler.InversionEnabler` class, which is itself a linear
289 290
operator.

Philipp Arras's avatar
Philipp Arras committed
291
.. currentmodule:: nifty5.operators.operator
Philipp Arras's avatar
Docs pa  
Philipp Arras committed
292

293
Direct multiplication and adjoint inverse multiplication transform a field
Philipp Arras's avatar
Philipp Arras committed
294 295 296 297
defined on the operator's :attr:`~Operator.domain` to one defined on the
operator's :attr:`~Operator.target`, whereas adjoint multiplication and inverse
multiplication transform from :attr:`~Operator.target` to
:attr:`~Operator.domain`.
Philipp Arras's avatar
Docs pa  
Philipp Arras committed
298 299

.. currentmodule:: nifty5.operators
300 301

Operators with identical domain and target can be derived from
Philipp Arras's avatar
Docs pa  
Philipp Arras committed
302 303 304 305 306 307 308
:class:`~endomorphic_operator.EndomorphicOperator`. Typical examples for this
category are the :class:`~scaling_operator.ScalingOperator`, which simply
multiplies its input by a scalar value, and
:class:`~diagonal_operator.DiagonalOperator`, which multiplies every value of
its input field with potentially different values.

.. currentmodule:: nifty5
309 310 311

Further operator classes provided by NIFTy are

Philipp Arras's avatar
Docs pa  
Philipp Arras committed
312 313 314 315 316 317 318 319 320
- :class:`~operators.harmonic_operators.HarmonicTransformOperator` for
  transforms from a harmonic domain to its counterpart in position space, and
  their adjoint
- :class:`~operators.distributors.PowerDistributor` for transforms from a
  :class:`~domains.power_space.PowerSpace` to an associated harmonic domain, and
  their adjoint.
- :class:`~operators.simple_linear_operators.GeometryRemover`, which transforms
  from structured domains to unstructured ones. This is typically needed when
  building instrument response operators.
321

Martin Reinecke's avatar
Martin Reinecke committed
322

Martin Reinecke's avatar
Martin Reinecke committed
323 324
Syntactic sugar
---------------
Martin Reinecke's avatar
Martin Reinecke committed
325

Martin Reinecke's avatar
Martin Reinecke committed
326
Nifty5 allows simple and intuitive construction of altered and combined
327
operators.
Philipp Arras's avatar
Docs pa  
Philipp Arras committed
328
As an example, if ``A``, ``B`` and ``C`` are of type :class:`~operators.linear_operator.LinearOperator`
Philipp Arras's avatar
Philipp Arras committed
329
and ``f1`` and ``f2`` are of type :class:`~field.Field`, writing::
330

Martin Reinecke's avatar
Martin Reinecke committed
331
    X = A(B.inverse(A.adjoint)) + C
332 333
    f2 = X(f1)

Philipp Arras's avatar
Docs pa  
Philipp Arras committed
334 335
.. currentmodule:: nifty5.operators.linear_operator

336 337 338 339 340
will perform the operation suggested intuitively by the notation, checking
domain compatibility while building the composed operator.
The properties :attr:`~LinearOperator.adjoint` and
:attr:`~LinearOperator.inverse` return a new operator which behaves as if it
were the original operator's adjoint or inverse, respectively.
341 342 343 344 345 346
The combined operator infers its domain and target from its constituents,
as well as the set of operations it can support.
Instantiating operator adjoints or inverses by :attr:`~LinearOperator.adjoint`
and similar methods is to be distinguished from the instant application of
operators performed by :attr:`~LinearOperator.adjoint_times` and similar
methods.
347 348 349 350


.. _minimization:

Martin Reinecke's avatar
Martin Reinecke committed
351

352 353 354
Minimization
============

Martin Reinecke's avatar
Martin Reinecke committed
355 356 357
Most problems in IFT are solved by (possibly nested) minimizations of
high-dimensional functions, which are often nonlinear.

Philipp Arras's avatar
Docs pa  
Philipp Arras committed
358
.. currentmodule:: nifty5.minimization
Martin Reinecke's avatar
Martin Reinecke committed
359 360 361

Energy functionals
------------------
362

Philipp Arras's avatar
Docs pa  
Philipp Arras committed
363 364 365 366 367
In NIFTy5 such functions are represented by objects of type
:class:`~energy.Energy`. These hold the prescription how to calculate the
function's :attr:`~energy.Energy.value`, :attr:`~energy.Energy.gradient` and
(optionally) :attr:`~energy.Energy.metric` at any given
:attr:`~energy.Energy.position` in parameter space. Function values are
368
floating-point scalars, gradients have the form of fields defined on the energy's
Philipp Arras's avatar
Docs pa  
Philipp Arras committed
369 370 371
position domain, and metrics are represented by linear operator objects.

.. currentmodule:: nifty5
372

Martin Reinecke's avatar
Martin Reinecke committed
373
Energies are classes that typically have to be provided by the user when
Philipp Arras's avatar
Docs pa  
Philipp Arras committed
374 375 376
tackling new IFT problems. An example of concrete energy classes delivered with
NIFTy5 is :class:`~minimization.quadratic_energy.QuadraticEnergy` (with
position-independent metric, mainly used with conjugate gradient minimization).
377

378 379 380 381 382 383 384 385 386 387 388 389
For MGVI, NIFTy provides the :class:`~energy.Energy` subclass :class:`~minimization.metric_gaussian_kl.MetricGaussianKL`,
which computes the sampled estimated of the KL divergence, its gradient and the Fisher metric. The constructor
of :class:`~minimization.metric_gaussian_kl.MetricGaussianKL` requires an instance of
:class:`~operators.energy_operators.StandardHamiltonian`, an operator to compute the negative log-likelihood of the problem in standardized coordinates
at a given position in parameter space. Finally, the :class:`~operators.energy_operators.StandardHamiltonian` can be constructed from
the likelihood, represented by an :class:`~operators.energy_operators.EnergyOperator` instance. Several commonly used forms of the likelihoods are already provided in
NIFTy, such as :class:`~operators.energy_operators.GaussianEnergy`, :class:`~operators.energy_operators.PoissonianEnergy`,
:class:`~operators.energy_operators.InverseGammaLikelihood` or :class:`~operators.energy_operators.BernoulliEnergy`, but the user
is free to implement a likelihood customized to the problem at hand. The dome code `demos/getting_started_3.py` illustrates how to set up an energy functional
for MGVI and minimize it.


Martin Reinecke's avatar
Martin Reinecke committed
390 391 392

Iteration control
-----------------
Martin Reinecke's avatar
Martin Reinecke committed
393

Philipp Arras's avatar
Docs pa  
Philipp Arras committed
394 395
.. currentmodule:: nifty5.minimization.iteration_controllers

396
Iterative minimization of an energy requires some means of
Martin Reinecke's avatar
Martin Reinecke committed
397
checking the quality of the current solution estimate and stopping once
Martin Reinecke's avatar
Martin Reinecke committed
398 399 400
it is sufficiently accurate. In case of numerical problems, the iteration needs
to be terminated as well, returning a suitable error description.

Philipp Arras's avatar
Philipp Arras committed
401
In NIFTy5, this functionality is encapsulated in the abstract
Martin Reinecke's avatar
Martin Reinecke committed
402 403 404 405 406 407 408
:class:`IterationController` class, which is provided with the initial energy
object before starting the minimization, and is updated with the improved
energy after every iteration. Based on this information, it can either continue
the minimization or return the current estimate indicating convergence or
failure.

Sensible stopping criteria can vary significantly with the problem being
409
solved; NIFTy provides a concrete sub-class of :class:`IterationController`
Martin Reinecke's avatar
Martin Reinecke committed
410
called :class:`GradientNormController`, which should be appropriate in many
411 412
circumstances. A full list of the available :class:`IterationController` s
in NIFTy can be found below, but users have complete freedom to implement custom
413
:class:`IterationController` sub-classes for their specific applications.
Martin Reinecke's avatar
Martin Reinecke committed
414 415 416 417

Minimization algorithms
-----------------------

Philipp Arras's avatar
Docs pa  
Philipp Arras committed
418 419
.. currentmodule:: nifty5.minimization

Martin Reinecke's avatar
Martin Reinecke committed
420
All minimization algorithms in NIFTy inherit from the abstract
Philipp Arras's avatar
Docs pa  
Philipp Arras committed
421
:class:`~minimizer.Minimizer` class, which presents a minimalistic interface
Philipp Arras's avatar
Philipp Arras committed
422
consisting only of a :meth:`~minimizer.Minimizer.__call__` method taking an
Philipp Arras's avatar
Docs pa  
Philipp Arras committed
423 424
:class:`~energy.Energy` object and optionally a preconditioning operator, and
returning the energy at the discovered minimum and a status code.
Martin Reinecke's avatar
Martin Reinecke committed
425

Philipp Arras's avatar
Docs pa  
Philipp Arras committed
426 427 428
For energies with a quadratic form (i.e. which can be expressed by means of a
:class:`~quadratic_energy.QuadraticEnergy` object), an obvious choice of
algorithm is the :class:`~conjugate_gradient.ConjugateGradient` minimizer.
Martin Reinecke's avatar
Martin Reinecke committed
429 430

A similar algorithm suited for nonlinear problems is provided by
Philipp Arras's avatar
Docs pa  
Philipp Arras committed
431
:class:`~nonlinear_cg.NonlinearCG`.
Martin Reinecke's avatar
Martin Reinecke committed
432

Martin Reinecke's avatar
Martin Reinecke committed
433
Many minimizers for nonlinear problems can be characterized as
Martin Reinecke's avatar
Martin Reinecke committed
434

Philipp Arras's avatar
Docs pa  
Philipp Arras committed
435 436
- First deciding on a direction for the next step.
- Then finding a suitable step length along this direction, resulting in the
Martin Reinecke's avatar
Martin Reinecke committed
437 438
  next energy estimate.

Philipp Arras's avatar
Docs pa  
Philipp Arras committed
439 440
This family of algorithms is encapsulated in NIFTy's
:class:`~descent_minimizers.DescentMinimizer` class, which currently has three
Martin Reinecke's avatar
fixes  
Martin Reinecke committed
441
generally usable concrete implementations:
Philipp Arras's avatar
Docs pa  
Philipp Arras committed
442 443
:class:`~descent_minimizers.NewtonCG`, :class:`~descent_minimizers.L_BFGS` and
:class:`~descent_minimizers.VL_BFGS`. Of these algorithms, only
Martin Reinecke's avatar
fixes  
Martin Reinecke committed
444
:class:`~descent_minimizers.NewtonCG` requires the energy object to provide
Philipp Arras's avatar
Docs pa  
Philipp Arras committed
445
a :attr:`~energy.Energy.metric` property, the others only need energy values and
446 447
gradients. Further available descent minimizers are :class:`~descent_minimizers.RelaxedNewton`
and :class:`~descent_minimizers.SteepestDescent`.
Philipp Arras's avatar
Docs pa  
Philipp Arras committed
448 449

The flexibility of NIFTy's design allows using externally provided minimizers.
450
With only small effort, adaptors for two SciPy minimizers were written; they are
Philipp Arras's avatar
Docs pa  
Philipp Arras committed
451
available under the names :class:`~scipy_minimizer.ScipyCG` and
Martin Reinecke's avatar
fixes  
Martin Reinecke committed
452
:class:`~scipy_minimizer.L_BFGS_B`.
Martin Reinecke's avatar
Martin Reinecke committed
453

Martin Reinecke's avatar
Martin Reinecke committed
454 455 456 457

Application to operator inversion
---------------------------------

Philipp Arras's avatar
Docs pa  
Philipp Arras committed
458
.. currentmodule:: nifty5
Martin Reinecke's avatar
Martin Reinecke committed
459

Philipp Arras's avatar
Docs pa  
Philipp Arras committed
460 461 462 463
The machinery presented here cannot only be used for minimizing functionals
derived from IFT, but also for the numerical inversion of linear operators, if
the desired application mode is not directly available. A classical example is
the information propagator whose inverse is defined as:
Martin Reinecke's avatar
Martin Reinecke committed
464

Martin Reinecke's avatar
Martin Reinecke committed
465
:math:`D^{-1} = \left(R^\dagger N^{-1} R + S^{-1}\right)`.
Martin Reinecke's avatar
Martin Reinecke committed
466

Philipp Arras's avatar
Docs pa  
Philipp Arras committed
467
It needs to be applied in forward direction in order to calculate the Wiener
468 469
filter solution, but only its inverse application is straightforward.
To use it in forward direction, we make use of NIFTy's
Philipp Arras's avatar
Docs pa  
Philipp Arras committed
470
:class:`~operators.inversion_enabler.InversionEnabler` class, which internally
Martin Reinecke's avatar
Martin Reinecke committed
471
applies the (approximate) inverse of the given operator :math:`x = Op^{-1} (y)` by
472 473 474 475 476
solving the equation :math:`y = Op (x)` for :math:`x`.
This is accomplished by minimizing a suitable
:class:`~minimization.quadratic_energy.QuadraticEnergy`
with the :class:`~minimization.conjugate_gradient.ConjugateGradient`
algorithm. An example is provided in
Philipp Arras's avatar
Philipp Arras committed
477
:func:`~library.wiener_filter_curvature.WienerFilterCurvature`.
478 479 480 481 482 483 484 485 486 487 488 489 490

      
Posterior analysis and visualization
---------------------------------

After the minimization of an energy functional has converged, samples can be drawn
from the posterior distribution at the current position to investigate the result.
The probing module offers class called :class:`~probing.StatCalculator`
which allows to evaluate the :attr:`~probing.StatCalculator.mean` and the unbiased
variance :attr:`~probing.StatCalculator.mean` of these samples.

Fields can be visualized using the :class:`~plot.Plot` class, which invokes
matplotlib for plotting.