ift.rst 15.7 KB
Newer Older
Martin Reinecke's avatar
Martin Reinecke committed
1
2
3
4
5
6
IFT -- Information Field Theory
===============================

Theoretical Background
----------------------

Martin Reinecke's avatar
Martin Reinecke committed
7
`Information Field Theory <https://www.mpa-garching.mpg.de/ift/>`_ [1]_  (IFT) is information theory, the logic of reasoning under uncertainty, applied to fields.
8
9
10
11
A field can be any quantity defined over some space, e.g. the air temperature over Europe, the magnetic field strength in the Milky Way, or the matter density in the Universe.
IFT describes how data and knowledge can be used to infer field properties.
Mathematically it is a statistical field theory and exploits many of the tools developed for such.
Practically, it is a framework for signal processing and image reconstruction.
Martin Reinecke's avatar
Martin Reinecke committed
12

13
14
IFT is fully Bayesian.
How else could infinitely many field degrees of freedom be constrained by finite data?
Martin Reinecke's avatar
Martin Reinecke committed
15

16
17
18
19
20
There is a full toolbox of methods that can be used, like the classical approximation (= Maximum a posteriori = MAP), effective action (= Variational Bayes = VI), Feynman diagrams, renormalization, and more.
IFT reproduces many known well working algorithms, which is reassuring.
Also, there were certainly previous works in a similar spirit.
Anyhow, in many cases IFT provides novel rigorous ways to extract information from data.
NIFTy comes with reimplemented MAP and VI estimators.
Martin Reinecke's avatar
Martin Reinecke committed
21

Martin Reinecke's avatar
Martin Reinecke committed
22
.. tip:: *In-a-nutshell introductions to information field theory* can be found in [2]_, [3]_, [4]_, and [5]_, with the latter probably being the most didactical.
Martin Reinecke's avatar
Martin Reinecke committed
23

Martin Reinecke's avatar
Martin Reinecke committed
24
.. [1] T.A. Enßlin et al. (2009), "Information field theory for cosmological perturbation reconstruction and nonlinear signal analysis", PhysRevD.80.105005, 09/2009; `[arXiv:0806.3474] <https://www.arxiv.org/abs/0806.3474>`_
Martin Reinecke's avatar
Martin Reinecke committed
25

Martin Reinecke's avatar
Martin Reinecke committed
26
.. [2] T.A. Enßlin (2013), "Information field theory", proceedings of MaxEnt 2012 -- the 32nd International Workshop on Bayesian Inference and Maximum Entropy Methods in Science and Engineering; AIP Conference Proceedings, Volume 1553, Issue 1, p.184; `[arXiv:1301.2556] <https://arxiv.org/abs/1301.2556>`_
27

Martin Reinecke's avatar
Martin Reinecke committed
28
.. [3] T.A. Enßlin (2014), "Astrophysical data analysis with information field theory", AIP Conference Proceedings, Volume 1636, Issue 1, p.49; `[arXiv:1405.7701] <https://arxiv.org/abs/1405.7701>`_
Martin Reinecke's avatar
Martin Reinecke committed
29

Martin Reinecke's avatar
cleanup    
Martin Reinecke committed
30
.. [4] Wikipedia contributors (2018), `"Information field theory" <https://en.wikipedia.org/w/index.php?title=Information_field_theory&oldid=876731720>`_, Wikipedia, The Free Encyclopedia.
Torsten Ensslin's avatar
Torsten Ensslin committed
31

Martin Reinecke's avatar
Martin Reinecke committed
32
.. [5] T.A. Enßlin (2019), "Information theory for fields", accepted by Annalen der Physik; `[DOI] <https://doi.org/10.1002/andp.201800127>`_, `[arXiv:1804.03350] <https://arxiv.org/abs/1804.03350>`_
33

Martin Reinecke's avatar
Martin Reinecke committed
34
35
36



Martin Reinecke's avatar
cleanup    
Martin Reinecke committed
37
Free Theory & Implicit Operators
38
39
--------------------------------

Martin Reinecke's avatar
Martin Reinecke committed
40
A free IFT appears when the signal field :math:`{s}` and the noise :math:`{n}` of the data :math:`{d}` are independent, zero-centered Gaussian processes of known covariances :math:`{S}` and :math:`{N}`, respectively,
41
42
43
44
45

.. math::

    \mathcal{P}(s,n) = \mathcal{G}(s,S)\,\mathcal{G}(n,N),

Martin Reinecke's avatar
cleanup    
Martin Reinecke committed
46
and the measurement equation is linear in both signal and noise,
47
48
49
50
51

.. math::

    d= R\, s + n,

Martin Reinecke's avatar
Martin Reinecke committed
52
with :math:`{R}` being the measurement response, which maps the continuous signal field into the discrete data space.
53
54

This is called a free theory, as the information Hamiltonian
Philipp Arras's avatar
Philipp Arras committed
55

56
57
58
59
.. math::

    \mathcal{H}(d,s)= -\log \mathcal{P}(d,s)= \frac{1}{2} s^\dagger S^{-1} s + \frac{1}{2} (d-R\,s)^\dagger N^{-1} (d-R\,s) + \mathrm{const}

Martin Reinecke's avatar
cleanup    
Martin Reinecke committed
60
is only of quadratic order in :math:`{s}`, which leads to a linear relation between the data and the posterior mean field.
61

Martin Reinecke's avatar
cleanup    
Martin Reinecke committed
62
In this case, the posterior is
63
64
65
66
67

.. math::

    \mathcal{P}(s|d) = \mathcal{G}(s-m,D)

Martin Reinecke's avatar
cleanup    
Martin Reinecke committed
68
with
69
70
71
72
73
74
75
76
77
78
79

.. math::

    m = D\, j

the posterior mean field,

.. math::

    D = \left( S^{-1} + R^\dagger N^{-1} R\right)^{-1}

Martin Reinecke's avatar
cleanup    
Martin Reinecke committed
80
the posterior covariance operator, and
81
82
83
84
85

.. math::

    j = R^\dagger N^{-1} d

86
87
the information source.
The operation in :math:`{m = D\,R^\dagger N^{-1} d}` is also called the generalized Wiener filter.
88

Martin Reinecke's avatar
Martin Reinecke committed
89
NIFTy permits to define the involved operators :math:`{R}`, :math:`{R^\dagger}`, :math:`{S}`, and :math:`{N}` implicitly, as routines that can be applied to vectors, but which do not require the explicit storage of the matrix elements of the operators.
Torsten Ensslin's avatar
Torsten Ensslin committed
90

91
92
93
Some of these operators are diagonal in harmonic (Fourier) basis, and therefore only require the specification of a (power) spectrum and :math:`{S= F\,\widehat{P_s} F^\dagger}`.
Here :math:`{F = \mathrm{HarmonicTransformOperator}}`, :math:`{\widehat{P_s} = \mathrm{DiagonalOperator}(P_s)}`, and :math:`{P_s(k)}` is the power spectrum of the process that generated :math:`{s}` as a function of the (absolute value of the) harmonic (Fourier) space coordinate :math:`{k}`.
For those, NIFTy can easily also provide inverse operators, as :math:`{S^{-1}= F\,\widehat{\frac{1}{P_s}} F^\dagger}` in case :math:`{F}` is unitary, :math:`{F^\dagger=F^{-1}}`.
94
95

These implicit operators can be combined into new operators, e.g. to :math:`{D^{-1} = S^{-1} + R^\dagger N^{-1} R}`, as well as their inverses, e.g. :math:`{D = \left( D^{-1} \right)^{-1}}`.
96
97
The invocation of an inverse operator applied to a vector might trigger the execution of a numerical linear algebra solver.

Martin Reinecke's avatar
Martin Reinecke committed
98
Thus, when NIFTy calculates :math:`{m = D\, j}`, it actually solves :math:`{D^{-1} m = j}` for :math:`{m}` behind the scenes.
Martin Reinecke's avatar
Martin Reinecke committed
99
100
The advantage of implicit operators compared to explicit matrices is the reduced memory consumption;
for the reconstruction of just a Megapixel image the latter would already require several Terabytes.
101
Larger images could not be dealt with due to the quadratic memory requirements of explicit operator representations.
102

Martin Reinecke's avatar
Martin Reinecke committed
103
The demo codes `demos/getting_started_1.py` and `demos/Wiener_Filter.ipynb` illustrate this.
104
105
106
107
108


Generative Models
-----------------

Martin Reinecke's avatar
Martin Reinecke committed
109
For more sophisticated measurement situations (involving non-linear measurements, unknown covariances, calibration constants and the like) it is recommended to formulate those as generative models for which NIFTy provides powerful inference algorithms.
110

Torsten Ensslin's avatar
Torsten Ensslin committed
111
In a generative model, all known or unknown quantities are described as the results of generative processes, which start with simple probability distributions, like the uniform, the i.i.d. Gaussian, or the delta distribution.
112

113
Let us rewrite the above free theory as a generative model:
Torsten Ensslin's avatar
Torsten Ensslin committed
114

115
116
117
.. math::

    s = A\,\xi
Torsten Ensslin's avatar
Torsten Ensslin committed
118

119
with :math:`{A}` the amplitude operator such that it generates signal field realizations with the correct covariance :math:`{S=A\,A^\dagger}` when being applied to a white Gaussian field :math:`{\xi}` with :math:`{\mathcal{P}(\xi)= \mathcal{G}(\xi, 1)}`.
120

121
The joint information Hamiltonian for the standardized signal field :math:`{\xi}` reads:
Torsten Ensslin's avatar
Torsten Ensslin committed
122

123
124
.. math::

125
    \mathcal{H}(d,\xi)= -\log \mathcal{P}(d,s)= \frac{1}{2} \xi^\dagger \xi + \frac{1}{2} (d-R\,A\,\xi)^\dagger N^{-1} (d-R\,A\,\xi) + \mathrm{const}.
126

Martin Reinecke's avatar
cleanup    
Martin Reinecke committed
127
NIFTy takes advantage of this formulation in several ways:
128

129
1) All prior degrees of freedom have unit covariance, which improves the condition number of operators that need to be inverted.
Martin Reinecke's avatar
Martin Reinecke committed
130

131
2) The amplitude operator can be regarded as part of the response, :math:`{R'=R\,A}`.
Martin Reinecke's avatar
Martin Reinecke committed
132
   In general, more sophisticated responses can be obtained by combining simpler operators.
Martin Reinecke's avatar
Martin Reinecke committed
133

Martin Reinecke's avatar
Martin Reinecke committed
134
3) The response can be non-linear, e.g. :math:`{R'(s)=R \exp(A\,\xi)}`, see `demos/getting_started_2.py`.
Martin Reinecke's avatar
Martin Reinecke committed
135

136
4) The amplitude operator may depend on further parameters, e.g. :math:`A=A(\tau)= F\, \widehat{e^\tau}` represents an amplitude operator with a positive definite, unknown spectrum defined in the Fourier domain.
Martin Reinecke's avatar
Martin Reinecke committed
137
138
   The amplitude field :math:`{\tau}` would get its own amplitude operator, with a cepstrum (spectrum of a log spectrum) defined in quefrency space (harmonic space of a logarithmically binned harmonic space) to regularize its degrees of freedom by imposing some (user-defined degree of) spectral smoothness.

139
5) NIFTy calculates the gradient of the information Hamiltonian and the Fisher information metric with respect to all unknown parameters, here :math:`{\xi}` and :math:`{\tau}`, by automatic differentiation.
Torsten Ensslin's avatar
Torsten Ensslin committed
140
   The gradients are used for MAP estimates, and the Fisher matrix is required in addition to the gradient by Metric Gaussian Variational Inference (MGVI), which is available in NIFTy as well.
Martin Reinecke's avatar
Martin Reinecke committed
141
   MGVI is an implicit operator extension of Automatic Differentiation Variational Inference (ADVI).
142

143
144
The reconstruction of a non-Gaussian signal with unknown covariance from a non-trivial (tomographic) response is demonstrated in `demos/getting_started_3.py`.
Here, the uncertainty of the field and the power spectrum of its generating process are probed via posterior samples provided by the MGVI algorithm.
145

Philipp Arras's avatar
Philipp Arras committed
146
147
148
149
+----------------------------------------------------+
| **Output of tomography demo getting_started_3.py** |
+----------------------------------------------------+
| .. image:: images/getting_started_3_setup.png      |
Philipp Arras's avatar
Philipp Arras committed
150
|                                                    |
Philipp Arras's avatar
Philipp Arras committed
151
152
153
154
155
156
+----------------------------------------------------+
| Non-Gaussian signal field,                         |
| data backprojected into the image domain, power    |
| spectrum of underlying Gausssian process.          |
+----------------------------------------------------+
| .. image:: images/getting_started_3_results.png    |
Philipp Arras's avatar
Philipp Arras committed
157
|                                                    |
Philipp Arras's avatar
Philipp Arras committed
158
159
160
161
162
163
164
+----------------------------------------------------+
| Posterior mean field signal                        |
| reconstruction, its uncertainty, and the power     |
| spectrum of the process for different posterior    |
| samples in comparison to the correct one (thick    |
| orange line).                                      |
+----------------------------------------------------+
165

166
Maximum a Posteriori
Torsten Ensslin's avatar
Torsten Ensslin committed
167
168
--------------------

Philipp Arras's avatar
Docs    
Philipp Arras committed
169
One popular field estimation method is Maximum a Posteriori (MAP).
170

Martin Reinecke's avatar
Martin Reinecke committed
171
It only requires minimizing the information Hamiltonian, e.g. by a gradient descent method that stops when
172
173
174
175
176

.. math::

    \frac{\partial \mathcal{H}(d,\xi)}{\partial \xi} = 0.

Martin Reinecke's avatar
Martin Reinecke committed
177
NIFTy7 automatically calculates the necessary gradient from a generative model of the signal and the data and uses this to minimize the Hamiltonian.
178

179
However, MAP often provides unsatisfactory results in cases of deep hirachical Bayesian networks.
180
181
182
183
184
185
The reason for this is that MAP ignores the volume factors in parameter space, which are not to be neglected in deciding whether a solution is reasonable or not.
In the high dimensional setting of field inference these volume factors can differ by large ratios.
A MAP estimate, which is only representative for a tiny fraction of the parameter space, might be a poorer choice (with respect to an error norm) compared to a slightly worse location with slightly lower posterior probability, which, however, is associated with a much larger volume (of nearby locations with similar probability).

This causes MAP signal estimates to be more prone to overfitting the noise as well as to perception thresholds than methods that take volume effects into account.

Torsten Ensslin's avatar
Torsten Ensslin committed
186

Philipp Frank's avatar
Philipp Frank committed
187
188
Metric Gaussian Variational Inference
-------------------------------------
Torsten Ensslin's avatar
Torsten Ensslin committed
189

Martin Reinecke's avatar
Martin Reinecke committed
190
One method that takes volume effects into account is Variational Inference (VI).
191
192
In VI, the posterior :math:`\mathcal{P}(\xi|d)` is approximated by a simpler, parametrized distribution, often a Gaussian :math:`\mathcal{Q}(\xi)=\mathcal{G}(\xi-m,D)`.
The parameters of :math:`\mathcal{Q}`, the mean :math:`m` and its covariance :math:`D` are obtained by minimization of an appropriate information distance measure between :math:`\mathcal{Q}` and :math:`\mathcal{P}`.
Martin Reinecke's avatar
Martin Reinecke committed
193
As a compromise between being optimal and being computationally affordable, the variational Kullback-Leibler (KL) divergence is used:
194
195
196
197
198
199
200

.. math::

    \mathrm{KL}(m,D|d)= \mathcal{D}_\mathrm{KL}(\mathcal{Q}||\mathcal{P})=
    \int \mathcal{D}\xi \,\mathcal{Q}(\xi) \log \left( \frac{\mathcal{Q}(\xi)}{\mathcal{P}(\xi)} \right)

Minimizing this with respect to all entries of the covariance :math:`D` is unfeasible for fields.
201
Therefore, Metric Gaussian Variational Inference (MGVI) approximates the posterior precision matrix :math:`D^{-1}` at the location of the current mean :math:`m` by the Bayesian Fisher information metric,
202
203
204

.. math::

205
    M \approx \left\langle \frac{\partial \mathcal{H}(d,\xi)}{\partial \xi} \, \frac{\partial \mathcal{H}(d,\xi)}{\partial \xi}^\dagger \right\rangle_{(d,\xi)}.
Torsten Ensslin's avatar
Torsten Ensslin committed
206

207
208
In practice the average is performed over :math:`\mathcal{P}(d,\xi)\approx \mathcal{P}(d|\xi)\,\delta(\xi-m)` by evaluating the expression at the current mean :math:`m`.
This results in a Fisher information metric of the likelihood evaluated at the mean plus the prior information metric.
209
210
Therefore we will only have to infer the mean of the approximate distribution.
The only term within the KL-divergence that explicitly depends on it is the Hamiltonian of the true problem averaged over the approximation:
Martin Reinecke's avatar
Martin Reinecke committed
211

Torsten Ensslin's avatar
Torsten Ensslin committed
212
213
.. math::

214
215
    \mathrm{KL}(m|d) \;\widehat{=}\;
    \left\langle  \mathcal{H}(\xi,d)    \right\rangle_{\mathcal{Q}(\xi)},
Torsten Ensslin's avatar
Torsten Ensslin committed
216

217
where :math:`\widehat{=}` expresses equality up to irrelvant (here not :math:`m`-dependent) terms.
Martin Reinecke's avatar
Martin Reinecke committed
218

Martin Reinecke's avatar
Martin Reinecke committed
219
Thus, only the gradient of the KL is needed with respect to this, which can be expressed as
Torsten Ensslin's avatar
Torsten Ensslin committed
220
221
222

.. math::

223
    \frac{\partial \mathrm{KL}(m|d)}{\partial m} = \left\langle \frac{\partial \mathcal{H}(d,\xi)}{\partial \xi}  \right\rangle_{\mathcal{G}(\xi-m,D)}.
Torsten Ensslin's avatar
Torsten Ensslin committed
224

225
226
We stochastically estimate the KL-divergence and gradients with a set of samples drawn from the approximate posterior distribution.
The particular structure of the covariance allows us to draw independent samples solving a certain system of equations.
Philipp Frank's avatar
Philipp Frank committed
227
228
This KL-divergence for MGVI is implemented by 
:func:`~nifty7.minimization.kl_energies.MetricGaussianKL` within NIFTy7.
Torsten Ensslin's avatar
Torsten Ensslin committed
229

Philipp Frank's avatar
Philipp Frank committed
230
It should be noted that MGVI can typically only provide a lower bound on the variance.
Torsten Ensslin's avatar
Torsten Ensslin committed
231

Philipp Frank's avatar
Philipp Frank committed
232
233
Geometric Variational Inference
-------------------------------
234

Philipp Frank's avatar
Philipp Frank committed
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
For non-linear posterior distributions :math:`\mathcal{P}(\xi|d)` an approximation with a Gaussian :math:`\mathcal{Q}(\xi)` in the coordinates :math:`\xi` is sub-optimal, as higher order interactions are ignored.
A better approximation can be achieved by constructing a coordinate system :math:`y = g\left(\xi\right)` in which the posterior is close to a Gaussian, and perform VI with a Gaussian :math:`\mathcal{Q}(y)` in these coordinates.

One useful coordinate system is obtained in case the metric :math:`M` of the posterior can be expressed as the pullback of the Euclidean metric using an invertible transformation :math:`g`:

.. math::

    M = \left(\frac{\partial g}{\partial \xi}\right)^T \frac{\partial g}{\partial \xi} \ .

In general, such a transformation can only be constructed locally, i.E. in a neighbourhood of some expansion point :math:`\bar{\xi}`, denoted as :math:`g_{\bar{\xi}}\left(\xi\right)`. Using :math:`g_{\bar{\xi}}`, the Geometric Variational Inference [6]_ (GeoVI) scheme uses a zero mean, unit Gaussian :math:`\mathcal{Q}(y) = \mathcal{G}(y, 1)` approximation, which can be expressed in :math:`\xi` coordinates via the pushforward using the inverse transformation :math:`\xi = g_{\bar{\xi}}^{-1}(y)`:

.. math::

    \mathcal{Q}_{\bar{\xi}}(\xi) = \left(g_{\bar{\xi}}^{-1} * \mathcal{Q}\right)(\xi) = \int \delta\left(\xi - g_{\bar{\xi}}^{-1}(y)\right) \ \mathcal{G}(y, 1) \ \mathcal{D}y \ ,
    
where :math:`\delta` denotes the kronecker-delta.

The remaining task in geoVI is to obtain the optimal expansion point :math:`\bar{\xi}` suth that :math:`\mathcal{Q}_{\bar{\xi}}` matches the posterior as good as possible. Analogous to the MGVI algorithm, :math:`\bar{\xi}` is obtained by minimization of the KL-divergenge between :math:`\mathcal{P}` and :math:`\mathcal{Q}_{\bar{\xi}}` w.r.t. :math:`\bar{\xi}`. Furthermore the KL is represented as a stochastic estimate using a set of samples drawn from :math:`\mathcal{Q}_{\bar{\xi}}` which is implemented in NIFTy7 via :func:`~nifty7.minimization.kl_energies.GeoMetricKL`.

A visual comparison of the MGVI and GeoVI algorithm can be found in the demo script `mgvi_visualized.py`.

.. [6] P. Frank, R. Leike, and T.A. Enßlin (2021), "Geometric Variational Inference"; `[arXiv:2105.10470] <https://arxiv.org/abs/2105.10470>`_