ift.rst 13.8 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
--------------------------------

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 kown 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 continous 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.
99 100 101
The advantage of implicit operators to explicit matrices is the reduced memory requirements.
The reconstruction of only a Megapixel image would otherwithe require the storage and processing of matrices with sizes of several Terabytes.
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
typo  
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
132 133
   In general, more sophisticated responses can be constructed out of the composition of simpler operators.

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).
170 171 172 173 174 175 176

It only requires to minimize the information Hamiltonian, e.g by a gradient descent method that stops when

.. math::

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

177
NIFTy5 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
186 187 188 189

Variational Inference
---------------------

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.
227
This KL-divergence for MGVI is implemented in the class :class:`~minimization.metric_gaussian_kl.MetricGaussianKL` within NIFTy5.
Torsten Ensslin's avatar
Torsten Ensslin committed
228 229


Martin Reinecke's avatar
Martin Reinecke committed
230
The demo `getting_started_3.py` for example not only infers a field this way, but also the power spectrum of the process that has generated the field.
Martin Reinecke's avatar
Martin Reinecke committed
231
The cross-correlation of field and power spectrum is taken care of in this process.
232 233
Posterior samples can be obtained to study this cross-correlation.

Martin Reinecke's avatar
Martin Reinecke committed
234
It should be noted that MGVI, as any VI method, can typically only provide a lower bound on the variance.