Kaczmarz method

The Kaczmarz method or Kaczmarz's algorithm is an iterative algorithm for solving linear equation systems . It was first discovered by the Polish mathematician Stefan Kaczmarz,[1] and was rediscovered in the field of image reconstruction from projections by Richard Gordon, Robert Bender, and Gabor Herman in 1970, where it is called the Algebraic Reconstruction Technique (ART).[2] ART includes the positivity constraint, making it nonlinear.[3]

The Kaczmarz method is applicable to any linear system of equations, but its computational advantage relative to other methods depends on the system being sparse. It has been demonstrated to be superior, in some biomedical imaging applications, to other methods such as the filtered backprojection method.[4]

It has many applications ranging from computed tomography (CT) to signal processing. It can be obtained also by applying to the hyperplanes, described by the linear system, the method of successive projections onto convex sets (POCS).[5][6]

Algorithm 1: Kaczmarz algorithm

Let be a linear system, let the number of rows of A, be the th row of complex-valued matrix , and let be arbitrary complex-valued initial approximation to the solution of . For compute:

where and denotes complex conjugation of .

If the linear system is consistent, converges to the minimum-norm solution, provided that the iterations start with the zero vector.

A more general algorithm can be defined using a relaxation parameter

There are versions of the method that converge to a regularized weighted least squares solution when applied to a system of inconsistent equations and, at least as far as initial behavior is concerned, at a lesser cost than other iterative methods, such as the conjugate gradient method.[7]

Algorithm 2: Randomized Kaczmarz algorithm

Recently, a randomized version of the Kaczmarz method for overdetermined linear systems was introduced by Thomas Strohmer and Roman Vershynin[8] in which the i-th equation is selected randomly with probability proportional to .

This method can be seen as a particular case of stochastic gradient descent .[9]

Under such circumstances converges exponentially fast to the solution of , and the rate of convergence depends only on the scaled condition number .

Theorem

Let be the solution of . Then Algorithm 1 converges to in expectation, with the average error:

Proof

We have

for all

Using the fact that we can write (1) as

for all

The main point of the proof is to view the left hand side in (2) as an expectation of some random variable. Namely, recall that the solution space of the equation of is the hyperplane , whose normal is Define a random vector Z whose values are the normals to all the equations of , with probabilities as in our algorithm:

with probability

Then (2) says that

for all

The orthogonal projection onto the solution space of a random equation of is given by

Now we are ready to analyze our algorithm. We want to show that the error reduces at each step in average (conditioned on the previous steps) by at least the factor of The next approximation is computed from as where are independent realizations of the random projection The vector is in the kernel of It is orthogonal to the solution space of the equation onto which projects, which contains the vector (recall that is the solution to all equations). The orthogonality of these two vectors then yields To complete the proof, we have to bound from below. By the definition of , we have

where are independent realizations of the random vector

Thus

Now we take the expectation of both sides conditional upon the choice of the random vectors (hence we fix the choice of the random projections and thus the random vectors and we average over the random vector ). Then

By (3) and the independence,

Taking the full expectation of both sides, we conclude that

The superiority of this selection was illustrated with the reconstruction of a bandlimited function from its nonuniformly spaced sampling values. However, it has been pointed out[10] that the reported success by Strohmer and Vershynin depends on the specific choices that were made there in translating the underlying problem, whose geometrical nature is to find a common point of a set of hyperplanes, into a system of algebraic equations. There will always be legitimate algebraic representations of the underlying problem for which the selection method in [8] will perform in an inferior manner.[8][10][11]

Algorithm 3: Gower-Richtarik algorithm

In 2015, Robert M. Gower and Peter Richtarik[12] developed a versatile randomized iterative method for solving a consistent system of linear equations which includes the randomized Kaczmarz algorithm as a special case. Other special cases include randomized coordinate descent, randomized Gaussian descent and randomized Newton method. Block versions and versions with importance sampling of all these methods also arise as special cases. The method is shown to enjoy exponential rate decay (in expectation) - also known as linear convergence, under very mild conditions on the way randomness enters the algorithm. The Gower-Richtarik method is the first algorithm uncovering a "sibling" relationship between these methods, some of which were independently proposed before, while many of which were new.

Insights about Randomized Kaczmarz

Interesting new insights about the randomized Kaczmarz method that can be gained from the analysis of the method include:

Six Equivalent Formulations

The Gower-Richtarik method enjoys six seemingly different but equivalent formulations, shedding additional light on how to interpret it (and, as a consequence, how to interpret its many variants, including randomized Kaczmarz):

We now describe some of these viewpoints. The method depends on 2 parameters:

1. Sketch and Project

Given previous iterate the new point is computed by drawing a random matrix (in an iid fashion from some fixed distribution), and setting

That is, is obtain as the projection of onto the randomly sketched system . The idea behind this method is to pick in such a way that a projection onto the sketched system is substantially simpler than the solution of the original system . Randomized Kaczmarz method is obtained by picking to be the identity matrix, and to be the unit coordinate vector with probability Different choices of and lead to different variants of the method.

2. Constrain and Approximate

A seemingly different but entirely equivalent formulation of the method (obtained via Lagrangian duality) is

where is also allowed to vary, and where is any solution of the system Hence, is obtained by first constraining the update to the linear subspace spanned by the columns of the random matrix , i.e., to

and then choosing the point from this subspace which best approximates . This formulation may look surprising as it seems impossible to perform the approximation step due to the fact that is not known (after all, this is what we are trying the compute!). However, it is still possible to do this, simply because computed this way is the same as computed via the sketch and project formulation and since does not appear there.

5. Random Update

The update can also be written explicitly as

where by we denote the Moore-Penrose pseudo-inverse of matrix . Hence, the method can be written in the form , where is a <bold>random update</bold> vector.

Letting it can be shown that the system always has a solution , and that for all such solutions the vector is the same. Hence, it does not matter which of these solutions is chosen, and the method can be also written as . The pseudo-inverse leads just to one particular solution. The role of the pseudo-inverse is twofold:

6. Random Fixed Point

If we subtract from both sides of the random update formula, denote and use the fact that we arrive at the last formulation:

where is the identity matrix. The iteration matrix, is random, whence the name of this formulation.

Convergence

By taking conditional expectations in the 6th formulation (conditional on ), we obtain

By taking expectation again, and using the tower property of expectations, we obtain

Gower and Richtarik [12] show that where the matrix norm is defined by Moreover, without any assumptions on one has By taking norms and unrolling the recurrence, we obtain

Theorem [Gower & Richtarik 2015]

Remark: A sufficient condition for the expected residuals to converge to 0 is This can be achieved if has a full column rank and under very mild conditions on Convergence of the method can be established also without the full column rank assumption in a different way.[13]

It is also possible to show a stronger result:

Theorem [Gower & Richtarik 2015]

The expected squared norms (rather than norms of expectations) converge at the same rate:

Remark: This second type of convergence is stronger due to the following identity [12] which holds for any random vector and any fixed vector :

Convergence of Randomized Kaczmarz

We have seen that the randomized Kaczmarz method appears as a special case of the Gower-Richtarik method for and being the unit coordinate vector with probability where is the row of It can be checked by direct calculation that

Notes

References

External links

This article is issued from Wikipedia - version of the 9/10/2016. The text is available under the Creative Commons Attribution/Share Alike but additional terms may apply for the media files.