TheKaczmarz method orKaczmarz's algorithm is aniterative algorithm for solvinglinear equation systems${\displaystyle Ax=b}$. It was first discovered by the Polish mathematicianStefan Kaczmarz,^[1] and was rediscovered in the field of image reconstruction from projections byRichard Gordon, Robert Bender, andGabor Herman in 1970, where it is called theAlgebraic 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 beingsparse. It has been demonstrated to be superior, in some biomedical imaging applications, to other methods such as thefiltered backprojection method.^[4]

It has many applications ranging fromcomputed tomography (CT) tosignal processing. It can be obtained also by applying to the hyperplanes, described by the linear system, the method of successiveprojections onto convex sets (POCS).^[5][6]

The original Kaczmarz algorithm solves a complex-valuedsystem of linear equations${\displaystyle Ax=b}$.

Let${\displaystyle a_i}$ be theconjugate transpose of the${\displaystyle i}$-th row of${\displaystyle A}$. Initialize${\displaystyle x_0}$ to be an arbitrary complex-valued initial approximation. (e.g.${\displaystyle x_0=0}$.) For${\displaystyle k=0,1,\dots }$ compute:

${\displaystyle x_{k+1}=x_k+\frac{b_{i_k}-⟨a_{i_k},x_k⟩}{\Vert a_{i_k}{\Vert }^2}{a}_{i_k}}$

1

where${\displaystyle i_0,i_1,i_2,\dots }$ iterates over the rows of${\displaystyle A}$ in any order, deterministic or random. It is only necessary that each row is iterated infinitely often.

When we are in the space of real vectors, the Kaczmarz iteration has a clear geometric meaning. It means projecting$x_k$ orthogonally to thehyperplane defined by$\{x:⟨a_i,x⟩=b_i\}$. In this interpretation, it is clear that if the Kaczmarz iteration converges, then it must converge to one of the solutions to$Ax=b$.

A more general algorithm can be defined using arelaxation parameter${\displaystyle {\lambda }^k}$

${\displaystyle x_{k+1}=x_k+{\lambda }_k\frac{b_{i_k}-⟨a_{i_k},x_k⟩}{\Vert a_{i_k}{\Vert }^2}{a}_{i_k}}$

If the system has a solution,${\displaystyle x_k}$ converges to the minimum-norm solution, provided that the iterations start with the zero vector. If the rows are iterated in order, and${\displaystyle {\lambda }_k=1}$, then convergence is exponential.

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 theconjugate gradient method.^[7]

In 2009, a randomized version of the Kaczmarz method foroverdetermined linear systems was introduced by Thomas Strohmer and Roman Vershynin^[8] in which thei-th equation is selected randomly with probability proportional to${\displaystyle \Vert a_i{\Vert }^2.}$

This method can be seen as a particular case ofstochastic gradient descent.^[9]

Under such circumstances${\displaystyle x_k}$ converges exponentially fast to the solution of${\displaystyle Ax=b,}$ and the rate of convergence depends only on the scaledcondition number${\displaystyle \kappa (A)}$.

Theorem. Let${\displaystyle x}$ be the solution of${\displaystyle Ax=b.}$ Then Algorithm 2 converges to${\displaystyle x}$ in expectation, with the average error:

${\displaystyle \mathbb{E}\Vert x_k-x{\Vert }^2\le {\left(1-\kappa (A{)}^{-2}\right)}^k\cdot \Vert x_0-x{\Vert }^2.}$

We have

${\displaystyle \mathrm{\forall }z\in {\mathbb{C}}^n:\sum _{j=1}^m|⟨z,a_j⟩{|}^2\ge \frac{\Vert z{\Vert }^2}{\Vert A^{-1}{\Vert }^2}}$

2

Using

${\displaystyle \Vert A{\Vert }^2=\sum _{j=1}^m\Vert a_j{\Vert }^2}$

we can write (2) as

${\displaystyle \mathrm{\forall }z\in {\mathbb{C}}^n:\sum _{j=1}^m\frac{\Vert a_j{\Vert }^2}{\Vert A{\Vert }^2}{\left|⟨z,\frac{a_j}{\Vert a_j\Vert }⟩\right|}^2\ge \kappa (A{)}^{-2}\Vert z{\Vert }^2}$

3

The main point of the proof is to view the left hand side in (3) as an expectation of some random variable. Namely, recall that the solution space of the${\displaystyle j-th}$ equation of${\displaystyle Ax=b}$ is the hyperplane

${\displaystyle \{y:⟨y,a_j⟩=b_j\},}$

whose normal is${\displaystyle \frac{a_j}{\Vert a_j{\Vert }^2}.}$ Define a random vectorZ whose values are the normals to all the equations of${\displaystyle Ax=b}$, with probabilities as in our algorithm:

${\displaystyle Z=\frac{a_j}{\Vert a_j\Vert }}$ with probability${\displaystyle \frac{\Vert a_j{\Vert }^2}{\Vert A{\Vert }^2}j=1,\dots ,m}$

Then (3) says that

${\displaystyle \mathrm{\forall }z\in {\mathbb{C}}^n:\mathbb{E}|⟨z,Z⟩{|}^2\ge \kappa (A{)}^{-2}\Vert z{\Vert }^2}$

4

The orthogonal projection${\displaystyle P}$ onto the solution space of a random equation of${\displaystyle Ax=b}$ is given by${\displaystyle Pz=z-⟨z-x,Z⟩Z.}$

Now we are ready to analyze our algorithm. We want to show that the error${\displaystyle \Vert x_k-x{\Vert }^2}$ reduces at each step in average (conditioned on the previous steps) by at least the factor of${\displaystyle (1-\kappa (A{)}^{-2}).}$ The next approximation${\displaystyle x_k}$ is computed from${\displaystyle x_{k-1}}$ as${\displaystyle x_k=P_k{x}_{k-1},}$ where${\displaystyle P_1,P_2,\dots }$ are independent realizations of the random projection${\displaystyle P.}$ The vector${\displaystyle x_{k-1}-x_k}$ is in the kernel of${\displaystyle P_k.}$ It is orthogonal to the solution space of the equation onto which${\displaystyle P_k}$ projects, which contains the vector${\displaystyle x_k-x}$ (recall that${\displaystyle x}$ is the solution to all equations). The orthogonality of these two vectors then yields

${\displaystyle \Vert x_k-x{\Vert }^2=\Vert x_{k-1}-x{\Vert }^2-\Vert x_{k-1}-x_k{\Vert }^2.}$

To complete the proof, we have to bound${\displaystyle \Vert x_{k-1}-x_k{\Vert }^2}$ from below. By the definition of${\displaystyle x_k}$, we have

${\displaystyle \Vert x_{k-1}-x_k\Vert =⟨x_{k-1}-x,Z_k⟩}$

where${\displaystyle Z_1,Z_2,\dots }$ are independent realizations of the random vector${\displaystyle Z.}$

Thus

${\displaystyle \Vert x_k-x{\Vert }^2\le \left(1-{\left|⟨\frac{x_{k-1}-x}{\Vert x_{k-1}-x\Vert },Z_k⟩\right|}^2\right)\Vert x_{k-1}-x{\Vert }^2.}$

Now we take the expectation of both sides conditional upon the choice of the random vectors${\displaystyle Z_1,\dots ,Z_{k-1}}$ (hence we fix the choice of the random projections${\displaystyle P_1,\dots ,P_{k-1}}$ and thus the random vectors${\displaystyle x_1,\dots ,x_{k-1}}$ and we average over the random vector${\displaystyle Z_k}$). Then

${\displaystyle {\mathbb{E}}_{Z_1,\dots ,Z_{k-1}}\Vert x_k-x{\Vert }^2=\left(1-{\mathbb{E}}_{Z_1,\dots ,Z_{k-1},Z_k}{\left|⟨\frac{x_{k-1}-x}{\Vert x_{k-1}-x\Vert },Z_k⟩\right|}^2\right)\Vert x_{k-1}-x{\Vert }^2.}$

By (4) and the independence,

${\displaystyle {\mathbb{E}}_{Z_1,\dots ,Z_{k-1}}\Vert x_k-x{\Vert }^2\le (1-\kappa (A{)}^{-2})\Vert x_{k-1}-x{\Vert }^2.}$

Taking the full expectation of both sides, we conclude that

${\displaystyle \mathbb{E}\Vert x_k-x{\Vert }^2\le (1-\kappa (A{)}^{-2})\mathbb{E}\Vert x_{k-1}-x{\Vert }^2.◼}$

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 tofind 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]

The Kaczmarz iteration (1) has a purely geometric interpretation: the algorithm successively projects the current iterate onto the hyperplane defined by the next equation. Hence, any scaling of the equations is irrelevant; it can also be seen from (1) that any (nonzero) scaling of the equations cancels out. Thus, in RK, one can use${\displaystyle \Vert a_i\Vert }$ or any other weights that may be relevant. Specifically, in the above-mentioned reconstruction example, the equations were chosen with probability proportional to the average distance of each sample point from its two nearest neighbors — a concept introduced byFeichtinger andGröchenig. For additional progress on this topic, see,^[12][13] and the references therein.

In 2015, Robert M. Gower andPeter Richtarik^[14] developed a versatile randomized iterative method for solving a consistent system of linear equations${\displaystyle Ax=b}$ which includes the randomized Kaczmarz algorithm as a special case. Other special cases include randomizedcoordinate descent, randomized Gaussian descent and randomized Newton method. Block versions and versions withimportance 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.

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

• The general rate of the Gower-Richtarik algorithm precisely recovers the rate of the randomized Kaczmarz method in the special case when it reduced to it.
• The choice of probabilities for which the randomized Kaczmarz algorithm was originally formulated and analyzed (probabilities proportional to the squares of the row norms) is not optimal. Optimal probabilities are the solution of a certain semidefinite program. The theoretical complexity of randomized Kaczmarz with the optimal probabilities can be arbitrarily better than the complexity for the standard probabilities. However, the amount by which it is better depends on the matrix${\displaystyle A}$. There are problems for which the standard probabilities are optimal.
• When applied to a system with matrix${\displaystyle A}$ which is positive definite, Randomized Kaczmarz method is equivalent to the Stochastic Gradient Descent (SGD) method (with a very special stepsize) for minimizing the strongly convex quadratic function${\displaystyle f(x)=\frac{1}{2}{x}^{T}Ax-b^{T}x.}$ Note that since${\displaystyle f}$ is convex, the minimizers of${\displaystyle f}$ must satisfy${\displaystyle \mathrm{\nabla }f(x)=0}$, which is equivalent to${\displaystyle Ax=b.}$ The "special stepsize" is the stepsize which leads to a point which in the one-dimensional line spanned by the stochastic gradient minimizes the Euclidean distance from the unknown(!) minimizer of${\displaystyle f}$, namely, from${\displaystyle x^{\ast }=A^{-1}b.}$ This insight is gained from a dual view of the iterative process (below described as "Optimization Viewpoint: Constrain and Approximate").

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):

• 1. Sketching viewpoint: Sketch & Project
• 2. Optimization viewpoint: Constrain and Approximate
• 3. Geometric viewpoint: Random Intersect
• 4. Algebraic viewpoint 1: Random Linear Solve
• 5. Algebraic viewpoint 2: Random Update
• 6. Analytic viewpoint: Random Fixed Point

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

• a positive definite matrix${\displaystyle B}$ giving rise to a weighted Euclidean inner product${\displaystyle ⟨x,y{⟩}_B:=x^{T}By}$ and the induced norm

${\displaystyle \Vert x{\Vert }_B={\left(⟨x,x{⟩}_B\right)}^{\frac{1}{2}},}$

• and a random matrix${\displaystyle S}$ with as many rows as${\displaystyle A}$ (and possibly random number of columns).

Given previous iterate${\displaystyle x^k,}$ the new point${\displaystyle x^{k+1}}$ is computed by drawing a random matrix${\displaystyle S}$ (in an iid fashion from some fixed distribution), and setting

${\displaystyle x^{k+1}=\underset{x}{\mathrm{a}\mathrm{r}\mathrm{g}\mathrm{m}\mathrm{i}\mathrm{n}}\Vert x-x^k{\Vert }_B\text{ subject to }{S}^{T}Ax=S^{T}b.}$

That is,${\displaystyle x^{k+1}}$ is obtained as the projection of${\displaystyle x^k}$ onto the randomly sketched system${\displaystyle S^{T}Ax=S^{T}b}$. The idea behind this method is to pick${\displaystyle S}$ in such a way that a projection onto the sketched system is substantially simpler than the solution of the original system${\displaystyle Ax=b}$. Randomized Kaczmarz method is obtained by picking${\displaystyle B}$ to be theidentity matrix, and${\displaystyle S}$ to be the${\displaystyle i^{th}}$ unitcoordinate vector with probability${\displaystyle p_i=\Vert a_i{\Vert }_2^2/\Vert A{\Vert }_F^2.}$ Different choices of${\displaystyle B}$ and${\displaystyle S}$ lead to different variants of the method.

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

${\displaystyle x^{k+1}=\underset{x}{\mathrm{a}\mathrm{r}\mathrm{g}\mathrm{m}\mathrm{i}\mathrm{n}}{\Vert x-x^{\ast }\Vert }_B\text{ subject to }x=x^k+B^{-1}{A}^{T}Sy,}$

where${\displaystyle y}$ is also allowed to vary, and where${\displaystyle x^{\ast }}$ is any solution of the system${\displaystyle Ax=b.}$ Hence,${\displaystyle x^{k+1}}$ is obtained by first constraining the update to the linear subspace spanned by the columns of the random matrix${\displaystyle B^{-1}{A}^{T}S}$, i.e., to

${\displaystyle \left\{h:h=B^{-1}{A}^{T}Sy,y\text{ can vary }\right\},}$

and then choosing the point${\displaystyle x}$ from this subspace which best approximates${\displaystyle x^{\ast }}$. This formulation may look surprising as it seems impossible to perform the approximation step due to the fact that${\displaystyle x^{\ast }}$ is not known (after all, this is what we are trying the compute!). However, it is still possible to do this, simply because${\displaystyle x^{k+1}}$ computed this way is the same as${\displaystyle x^{k+1}}$ computed via the sketch and project formulation and since${\displaystyle x^{\ast }}$ does not appear there.

The update can also be written explicitly as

${\displaystyle x^{k+1}=x^k-B^{-1}{A}^{T}S{\left(S^{T}A{B}^{-1}{A}^{T}S\right)}^{†}{S}^T\left(A{x}^k-b\right),}$

where by${\displaystyle M^{†}}$ we denote the Moore-Penrose pseudo-inverse of matrix${\displaystyle M}$. Hence, the method can be written in the form${\displaystyle x^{k+1}=x^k+h^k}$, where${\displaystyle h^k}$ is arandom update vector.

Letting${\displaystyle M=S^{T}A{B}^{-1}{A}^{T}S,}$ it can be shown that the system${\displaystyle My=S^T(A{x}^k-b)}$ always has a solution${\displaystyle y^k}$, and that for all such solutions the vector${\displaystyle x^{k+1}-B^{-1}{A}^{T}S{y}^k}$ is the same. Hence, it does not matter which of these solutions is chosen, and the method can be also written as${\displaystyle x^{k+1}=x^k-B^{-1}{A}^{T}S{y}^k}$. The pseudo-inverse leads just to one particular solution. The role of the pseudo-inverse is twofold:

• It allows the method to be written in the explicit "random update" form as above,
• It makes the analysis simple through the final, sixth, formulation.

If we subtract${\displaystyle x^{\ast }}$ from both sides of the random update formula, denote

${\displaystyle Z:=A^{T}S{\left(S^{T}A{B}^{-1}{A}^{T}S\right)}^{†}{S}^{T}A,}$

and use the fact that${\displaystyle A{x}^{\ast }=b,}$ we arrive at the last formulation:

${\displaystyle x^{k+1}-x^{\ast }=\left(I-B^{-1}Z\right)\left(x^k-x^{\ast }\right),}$

where${\displaystyle I}$ is the identity matrix. The iteration matrix,${\displaystyle I-B^{-1}Z,}$ is random, whence the name of this formulation.

By taking conditional expectations in the 6th formulation (conditional on${\displaystyle x^k}$), we obtain

${\displaystyle \mathbb{E}\left[x^{k+1}-x^{\ast }\right|x^k]=\left(I-B^{-1}\mathbb{E}[Z]\right)\left[x^k-x^{\ast }\right].}$

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

${\displaystyle \mathbb{E}\left[x^{k+1}-x^{\ast }\right]=(I-B^{-1}\mathbb{E}[Z])\mathbb{E}\left[x^k-x^{\ast }\right].}$

Gower and Richtarik^[14] show that

${\displaystyle \rho :={\Vert I-B^{-\frac{1}{2}}\mathbb{E}[Z]B^{-\frac{1}{2}}\Vert }_B={\lambda }_{max}\left(I-B^{-1}\mathbb{E}[Z]\right),}$

where the matrix norm is defined by

${\displaystyle \Vert M{\Vert }_B:=\underset{x\ne 0}{max}\frac{\Vert Mx{\Vert }_B}{\Vert x{\Vert }_B}.}$

Moreover, without any assumptions on${\displaystyle S}$ one has${\displaystyle 0\le \rho \le 1.}$ By taking norms and unrolling the recurrence, we obtain

${\displaystyle {\Vert \mathbb{E}\left[x^k-x^{\ast }\right]\Vert }_B\le {\rho }^k\Vert x^0-x^{\ast }{\Vert }_B.}$

Remark. A sufficient condition for the expected residuals to converge to 0 is This can be achieved if${\displaystyle A}$ has a full column rank and under very mild conditions on${\displaystyle S.}$ Convergence of the method can be established also without the full column rank assumption in a different way.^[15]

It is also possible to show a stronger result:

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

${\displaystyle \mathbb{E}{\Vert \left[x^k-x^{\ast }\right]\Vert }_B^2\le {\rho }^k{\Vert x^0-x^{\ast }\Vert }_B^2.}$

Remark. This second type of convergence isstronger due to the following identity^[14] which holds for any random vector${\displaystyle x}$ and any fixed vector${\displaystyle x^{\ast }}$:

${\displaystyle {\Vert \mathbb{E}\left[x-x^{\ast }\right]\Vert }^2=\mathbb{E}\left[{\Vert x-x^{\ast }\Vert }^2\right]-\mathbb{E}\left[\Vert x-\mathbb{E}[x]{\Vert }^2\right].}$

We have seen that the randomized Kaczmarz method appears as a special case of the Gower-Richtarik method for${\displaystyle B=I}$ and${\displaystyle S}$ being the${\displaystyle i^{th}}$ unit coordinate vector with probability${\displaystyle p_i=\Vert a_i{\Vert }_2^2/\Vert A{\Vert }_F^2,}$ where${\displaystyle a_i}$ is the${\displaystyle i^{th}}$ row of${\displaystyle A.}$ It can be checked by direct calculation that

${\displaystyle \rho =\Vert I-B^{-1}\mathbb{E}[Z]{\Vert }_B=1-\frac{{\lambda }_{min}(A^{T}A)}{\Vert A{\Vert }_F^2}.}$

Since the convergence of the (randomized) Kaczmarz method depends on arate of convergence the method may make slow progress on some practical problems.^[10] To ensure finite termination of the method, Johannes Brust andMichael Saunders (academic)^[16] have developed a process that generalizes the (randomized) Kaczmarz iteration andterminates in at most${\displaystyle m}$ iterations to a solution for the consistent system${\displaystyle Ax=b}$. The process is based onDimensionality reduction, or projectionsonto lower dimensional spaces, which is how it derives its name PLSS (Projected Linear Systems Solver). An iteration of PLSS-Kaczmarz can be regarded as the generalization

${\displaystyle x^{k+1}=x^k+A_{:,1:k}^T(A_{1:k,:}{A}_{:,1:k}^T{)}^{†}(b_{1:k}-A_{1:k,:}{x}^k)}$

where${\displaystyle A_{1:k,:}}$ is the selection of rows 1 to${\displaystyle k}$ and all columns of${\displaystyle A}$. A randomized version of the method uses${\displaystyle k}$non repeated row indices at each iteration:${\displaystyle \{i_1,\dots ,i_{k-1},i_k\}}$ where each${\displaystyle i_j}$ is in${\displaystyle 1,2,...,m}$.The iteration converges to a solution when${\displaystyle k=m}$. In particular, since${\displaystyle A_{1:m,:}=A}$ it holds that

${\displaystyle A{x}^{m+1}=A{x}^m+A{A}^T(A{A}^T{)}^{†}(b-A{x}^m)=b}$

and therefore${\displaystyle x^{m+1}}$ is a solution to the linear system. The computation of iterates in PLSS-Kaczmarz can be simplified and organized effectively.The resulting algorithm only requires matrix-vector products and has a direct form

algorithm PLSS-Kaczmarzisinput: matrixA right hand sideboutput: solutionx such thatAx=bx := 0,P = [0]forkin1,2,...,mdoa :=A(ik,:)'                   // Select an index ik in 1,...,m without resamplingd :=P' * ac1 :=norm(a)c2 :=norm(d)c3 :=(bik-x'*a)/((c1-c2)*(c1+c2))p := c3*(a - P*(P'*a))P := [ P, p/norm(p) ]           // Append a normalized updatex := x + preturnx

• Kaczmarz, Stefan (1937),"Angenäherte Auflösung von Systemen linearer Gleichungen"(PDF),Bulletin International de l'Académie Polonaise des Sciences et des Lettres. Classe des Sciences Mathématiques et Naturelles. Série A, Sciences Mathématiques, vol. 35, pp. 355–357, archived fromthe original(PDF) on 2012-04-25, retrieved2011-10-07
• Chong, Edwin K. P.; Zak, Stanislaw H. (2008),An Introduction to Optimization (3rd ed.), John Wiley & Sons, pp. 226–230
• Gordon, Richard;Bender, Robert;Herman, Gabor (1970), "Algebraic reconstruction techniques (ART) for threedimensional electron microscopy and x-ray photography",Journal of Theoretical Biology,29 (3):471–481,Bibcode:1970JThBi..29..471G,doi:10.1016/0022-5193(70)90109-8,PMID 5492997
• Gordon, Richard (2011),Stop breast cancer now! Imagining imaging pathways towards search, destroy, cure and watchful waiting of premetastasis breast cancer. In: Breast Cancer - A Lobar Disease, editor: Tibor Tot, Springer, pp. 167–203
• Herman, Gabor (2009),Fundamentals of computerized tomography: Image reconstruction from projection (2nd ed.), Springer,ISBN 9781846287237
• Censor, Yair; Zenios, S.A. (1997),Parallel optimization: theory, algorithms, and applications, New York: Oxford University Press
• Aster, Richard; Borchers, Brian; Thurber, Clifford (2004),Parameter Estimation and Inverse Problems, Elsevier
• Strohmer, Thomas; Vershynin, Roman (2009),"A randomized Kaczmarz algorithm for linear systems with exponential convergence"(PDF),Journal of Fourier Analysis and Applications,15 (2):262–278,arXiv:math/0702226,doi:10.1007/s00041-008-9030-4,S2CID 1903919
• Needell, Deanna; Srebro, Nati; Ward, Rachel (2015), "Stochastic gradient descent, weighted sampling, and the randomized Kaczmarz algorithm",Mathematical Programming,155 (1–2):549–573,arXiv:1310.5715,doi:10.1007/s10107-015-0864-7,S2CID 2370209
• Censor, Yair;Herman, Gabor; Jiang, M. (2009), "A note on the behavior of the randomized Kaczmarz algorithm of Strohmer and Vershynin",Journal of Fourier Analysis and Applications,15 (4):431–436,Bibcode:2009JFAA...15..431C,doi:10.1007/s00041-009-9077-x,PMC 2872793,PMID 20495623
• Strohmer, Thomas; Vershynin, Roman (2009b), "Comments on the randomized Kaczmarz method",Journal of Fourier Analysis and Applications,15 (4):437–440,Bibcode:2009JFAA...15..437S,doi:10.1007/s00041-009-9082-0,S2CID 14806325
• Bass, Richard F.; Gröchenig, Karlheinz (2013), "Relevant sampling of band-limited functions",Illinois Journal of Mathematics,57 (1):43–58,arXiv:1203.0146,doi:10.1215/ijm/1403534485,S2CID 42705738
• Gordon, Dan (2017), "A derandomization approach to recovering bandlimited signals across a wide range of random sampling rates",Numerical Algorithms,77 (4):1141–1157,doi:10.1007/s11075-017-0356-3,S2CID 1794974
• Vinh Nguyen, Quang; Lumban Gaol, Ford (2011),Proceedings of the 2011 2nd International Congress on Computer Applications and Computational Science, vol. 2, Springer, pp. 465–469
• Gower, Robert;Richtarik, Peter (2015a), "Randomized iterative methods for linear systems",SIAM Journal on Matrix Analysis and Applications,36 (4):1660–1690,arXiv:1506.03296,doi:10.1137/15M1025487,S2CID 8215294
• Gower, Robert;Richtarik, Peter (2015b), "Stochastic dual ascent for solving linear systems",arXiv:1512.06890 [math.NA]
• Brust, Johannes J;Saunders, Michael A (2023), "PLSS: A Projected Linear Systems Solver",SIAM Journal on Scientific Computing,45 (2):A1012–A1037,arXiv:2207.07615,Bibcode:2023SJSC...45A1012B,doi:10.1137/22M1509783

• [1] A randomized Kaczmarz algorithm with exponential convergence
• [2]^{[permanent dead link]} Comments on the randomized Kaczmarz method
• [3] Kaczmarz algorithm in training Kolmogorov-Arnold network

• Numerical linear algebra
• Medical imaging
• Signal processing

• Articles with short description
• Short description matches Wikidata
• All articles with dead external links
• Articles with dead external links from July 2025
• Articles with permanently dead external links