Concept

Krylov subspace — where it appears

The span of a vector and its images under repeated multiplication by a matrix, which is the space every method in this family searches. It is built from products with the matrix alone, which is why every method in the family survives when the matrix is a subroutine rather than an array.

Named by 39 essays across 11 fields — each of them below, with the objects they name alongside it.

015304560759010512010⁻²10⁻¹110¹steprelative sizeleast error: 20discrepancy stop: 7errorresidualthe knob is an integerleast error, at step20error there0.14error at step 1206the residual falls at every stepthe error turns and keeps rising

A parameter that counts steps

The regularisation field's knob is a positive real number chosen by one of three rules. The iterative field's is an integer nobody called a knob — where to stop. On the same problem the best step is 20 and the best λ is 0.025, and they reach 0.1426 and 0.1406.

combination · Iterative regularisation
0408012016020024010⁻¹⁶10⁻¹³10⁻¹⁰10⁻⁷10⁻⁴10⁻¹iteration‖e‖ ⁄ ‖e₀‖ in the A-normmeasuredκ bound119 steps40×40, spectrum spread evenly in logbound permits 1417

The rate the condition number predicts

Conjugate gradients converge at a rate governed by the square root of the condition number. That is a bound rather than an estimate, it is provable, and it is loose enough that provisioning iterations from it wastes nine out of ten.

iterative · Krylov
10⁻¹⁵10⁻¹²10⁻⁹10⁻⁶10⁻³110⁻⁴10⁻³10⁻²10⁻¹1inner tolerance η, relative residual of the linear solvedistance from the root after the stepd² = 0.00138distance before the step, 0.03721093547231126the number by each point is the iterations it costeleven decades, one landing placedistance before the step0.037its square0.0014where η = 10⁻³ lands0.0025where η = 10⁻¹⁴ lands0.0025iterations for the first354iterations for the second1126the accuracy that is thrown awaymeasured against a root that is known

The accuracy that is thrown away

A Newton step is the exact answer to a linearised problem, and the linearisation is wrong at second order. So there is a floor under how close the step can land, the floor is the square of where it started, and eleven decades of inner tolerance below it buy the same four digits at four times the price.

sequence · Inexact newton
05101520253010⁻¹⁷10⁻¹⁴10⁻¹¹10⁻⁸10⁻⁵10⁻²10¹iteration‖RᵀR − I‖ and ‖r‖/‖b‖step n‖RᵀR − I‖residual30×30, run for exactly n stepsexact arithmetic would end here

An orthogonalisation nobody calls one

Conjugate gradients are derived as a minimisation and behave as an orthogonalisation, which is why the finite-termination property in every textbook is not a property the method has in floating point.

iterative · Krylov
0123456789101110⁻¹⁵10⁻¹²10⁻⁹10⁻⁶10⁻³110³Newton steptolerance asked for, and iterations paiditerations paidtolerance asked forouter residualthe adaptive policy, step by stepNewton steps10inner iterations, total1009first step's cost1last step's cost271final outer residual3.4·10⁻¹¹the rule reads the last two residualsand asks for nothing it cannot use

A tolerance that reads its own residual

The cheapest constant forcing term costs 980 inner iterations and arrives with a hundred times the forward error of the dearest, which costs 9,358. A rule that sets each step's tolerance from the ratio of the last two residuals costs 1,009 and arrives with neither problem — and it is not a constant, so it does not appear on the curve the constants are compared on.

sequence · Inexact newton
00.3670080.7340171.101031.468031.835040eigenvalue of P⁻¹Kwritten down, then computeddistinct3at 16φ computed1.6off the closed form2.9·10⁻¹⁴1 − φ1φthe preconditioner's effect is a theoremand the golden ratio is in it

Three eigenvalues, and two are the golden ratio

Precondition a saddle-point system by the block diagonal of its own two definite pieces and the preconditioned matrix has exactly three distinct eigenvalues — 1, and the two roots of λ² − λ − 1. A minimal polynomial of degree three means three steps, at every conditioning, and the preconditioner nobody can afford turns out to be the statement the affordable ones are measured against.

constraint · Block preconditioning
036912151821242700.250.50.7511.25index kweight appliedonebidiagonalArnoldiis it a function of σmisfit, even fit1.6·10⁻¹²misfit, general fit0.063‖A − Aᵀ‖/‖A‖0.086one method's weights do not notice the operatorand the other's stop being a function of σ

The basis decides what a filter is

The vocabulary of regularisation is spectral — a method keeps a component or discards it, and the weights are a function of the singular value. Row-normalising a symmetric blur so that it preserves a constant makes it 8.6% asymmetric, and that is enough to move GMRES's weights from 7·10⁻¹⁴ off a function of σ to 4.4·10⁻².

regularisation · GMRES
10⁻¹110¹10²10³10⁻¹⁷10⁻¹⁴10⁻¹¹10⁻⁸10⁻⁵s, on the real axis|H − Hᵣ| ÷ |H|exact where askedpoints4conditions bought8worst at a point5.3·10⁻¹⁶worst away from one8.4·10⁻⁴4 points, 8 conditionsand no bound in between

Exact at the points that were named

Balanced truncation asks for nothing and bounds everything, at a cost no large model can pay. The other kind of reduction asks for r numbers, costs r solves, is exact at every one of them — and bounds nothing anywhere else. That trade is the whole of large-scale model reduction.

reduction · Moment matching
02468101210⁻¹⁴10⁻¹²10⁻¹⁰10⁻⁸10⁻⁶10⁻⁴10⁻²1iteration‖r‖ / ‖b‖Laplaciancyclic shiftno progress at allevery eigenvalue of the shift is on the unit circleand it predicts nothing

The spectrum that predicts nothing

For a symmetric matrix the eigenvalues govern how fast an iteration converges. Drop symmetry and they stop governing anything — there is a matrix whose eigenvalues are as evenly spread as eigenvalues can be, on which GMRES makes no progress at all until the last possible step.

iterative · GMRES
10²10⁴10⁶10⁸10¹⁰110²10⁴10⁶10⁸10¹⁰condition number of the matrixcondition number seen by the iterationκ(A), unchangedκ(AR⁻¹)a bound with no κ(A) in itκ(AR⁻¹), every κ(A)2.2κ(SU), the other route2.2κ(A), across the sweep10⁸κ(AR⁻¹), across the sweep1the sketch never sees the spectrumand the spectrum cancels out of the answer

The sketch that is not the answer

Sketch-and-solve throws away the original problem and keeps the small one's answer, which is why its answer moves with the seed. Use the same sketch as a preconditioner instead and the condition number the iteration sees is the same number at every κ from a hundred to ten billion — identically the same, to nine digits, because the spectrum cancels out of it.

randomised · Sketching
2345678110²10⁴10⁶10⁸10¹⁰vectors in the basisκ₂ of the basisas derivedsolves spread outone subspace, two spanning setseight moments at one point7.7·10⁹eight points, spread1growth per vector661/u4.5·10¹⁵the same subspaceand only one of them usable

A basis that is the same subspace and not the same thing

The interpolation conditions are conditions on a subspace, so any basis of it will do. The one a derivation writes down reaches a condition number of 7.7·10⁹ by its eighth vector, and the rate at which it gets there is set by a number the user chose with no information.

reduction · Moment matching
015030045060075090010⁻¹³10⁻¹¹10⁻⁹10⁻⁷10⁻⁵10⁻³10⁻¹iteration‖r‖ / ‖b‖JacobiGauss–SeidelSOR ω=1.78closed form vs measuredρ Jacobi, exact0.99ρ measured0.99ρ Gauss–Seidel, exact0.981D Laplacian, n = 24ω optimal = 1.777

A rate that is known in advance

On the model problem, Jacobi contracts by cos(π/(n+1)) per step, Gauss–Seidel by its square, and optimally relaxed SOR by a number given in closed form. Three rates, all known before anything runs, and all measurable against what runs.

iterative · Stationary
26101418222630343810⁻¹⁷10⁻¹⁴10⁻¹¹10⁻⁸10⁻⁵10⁻²10¹stepsizeleast error: 20‖QᵀQ − I‖filter disagreementan identity with an expiry datedisagreement at step 1210⁻¹³disagreement at step 400.014least error at step20exact while the basis is orthogonaland false where the method is best

An expiry date the noise does not move

The polynomial description of conjugate gradients leaves the level of rounding at step 17 or 18 on this operator, at every noise level from 10% to 0.1%. The step worth stopping at moves from 3 to 44 across the same range. They coincide at about 1% noise, which is where the coincidence was first read, and it is a fact about the noise rather than about the method.

combination · Iterative regularisation
-1-0.500.511.50eigenvalue of P⁻¹Ktriangular [[H, Aᵀ], [0, −Ŝ]] — 1 valuediagonal blkdiag(H, Ŝ) — 3 valuessteps to a residual of 10⁻¹⁰GMRES, triangular2MINRES, diagonal3‖P⁻¹K − I‖54computed |λ − 1| at c = 18.1·10⁻⁸one copy of each value against twoand the counts follow

One eigenvalue and two steps

Put the off-diagonal block back into a block-diagonal saddle-point preconditioner and every eigenvalue of the preconditioned matrix becomes exactly one. GMRES still needs two steps, because the matrix is the identity plus a nilpotent part of norm 54, and a computed eigenvalue at one comes back as a ring of radius 8·10⁻⁸ — the square root of the rounding, not the rounding. With an approximate Schur complement the triangular form leaves one copy of each value where the diagonal form leaves two, and the step count halves.

constraint · Block preconditioning
11.31.61.92.210⁻¹⁶10⁻¹³10⁻¹⁰10⁻⁷10⁻⁴10⁻¹10²log₁₀ quadrature pointsdistance from the true counthalf an eigenvaluean integer, eventuallytrue count2at 4 points2at 128 points2finest error1.4·10⁻¹³the integral is an integerand a rounding hides how far it was

The last digit is the cheapest

Every cost curve measured here has the same shape: the first digits are cheap and the last ones are not. One method inverts it. Doubling the work buys twice as many digits as the previous doubling did, so the price of a digit halves every time it is paid.

cost · Nonlinear cost
10¹10³10⁵10⁷10⁹10¹¹10¹³10⁻¹⁷10⁻¹⁴10⁻¹¹10⁻⁸10⁻⁵10⁻²10¹condition number κ(A)‖QᵀQ − I‖classical once, Householdermodified once, Householderclassical twice, Householderclassical twice, Cholesky QRκ²uκuat κ = 10⁸classical once, Householder0.0042modified once, Householder5.7·10⁻⁹classical twice, Householder3.1·10⁻¹⁵classical twice, Cholesky QR10⁻¹⁵64×16 in blocks of 4, three seedsHouseholder inside does not help between

A stable block is not a stable basis

Block Gram–Schmidt orthogonalises twice over — between blocks, and inside each one. Householder inside the blocks does not stop the classical between-block step losing orthogonality like κ², 4.2·10⁻³ at κ = 4.3·10⁷, and a second pass does not stop Cholesky QR inside the blocks breaking down at κ = 10⁸. Each level fails only on ill-conditioning placed at its own level, and one variant holds 3·10⁻¹⁵ on every placement.

orthogonality · Gram–Schmidt
conjugate gradientsbest step20best error0.1410% window, last/first6.5Landweberbest step1778best error0.1410% window, last/first901110¹10²10³10⁴10⁵10⁶10⁷10⁻¹110¹10²matrix–vector productsrelative errorCGLS best: step 20Landweber best: step 1,778CGLSLandweberthe same answer at two pricesand a window three orders wide

A step that is not a unit of work

Landweber's iteration reaches conjugate gradients' best answer on the same deconvolution — 0.1414 against 0.1426 — at step 1,778 instead of step 20, and at 0.1% noise at step 56,234 instead of 44. Each step costs the same two products. And within 10% of its best it runs from step 7 to step 6,310, where conjugate gradients runs from 4 to 26: the slow method is the one that forgives a late stop.

combination · Iterative regularisation
significand bits10³10⁴10⁵10⁶10⁷10⁸10⁹10¹⁰κ(A)87892634456329625911109435761199127514162024every matrix positive definiteruns producing a false certificate14of runs in total72never above, in significand bits12first κ at eight bits10⁵the comparison was correctand what it proved was not true

Deciding that a zero has arrived

The previous tolerances were offers — accept this much error, save this much work. A detection threshold is not an offer, because both directions are failures. One matrix here has three genuinely near-invariant subspaces, and the constant somebody typed decides which of them the recurrence stops at; at eight significand bits the same kind of constant produces a proof of something false.

error · Deliberate zero
1357910⁻⁵10⁻⁴10⁻³10⁻²10⁻¹1eigenvalue λamplificationkeptdiscardeda filter on the starting vectormeasured against the polynomial3.9·10⁻¹¹worst kept direction1best discarded direction0.8the roots are the discarded Ritz valuesand the vertical lines are where they sit

Restarting is a filter

A restart throws away the Ritz values it does not want and begins again from a new starting vector. Written in the eigenbasis, that vector's components have been multiplied by a polynomial with its roots at the discarded values — measured component by component, and agreeing with the polynomial to rounding.

spectra · Lanczos
12345610⁻¹⁷10⁻¹⁴10⁻¹¹10⁻⁸10⁻⁵10⁻²10¹Ritz value, largest firstdistance from 10one vectora block of twoa space, not a ratecopies found, one vector1copies found, block of two2Krylov dimension, one vector16the second copy is not in the spaceat any number of steps

An eigenvalue one vector cannot see

A matrix with an exactly doubled eigenvalue at 10. Twelve Lanczos steps find it once; twenty-four find it once, on a Krylov space of dimension 23 in a 24-dimensional problem. A block of two vectors finds it twice. This is not slow convergence — the second copy is not in the space.

spectra · Invariant subspace
total inner iterationscold at η = 1e-107016scaled warm start6837unscaled728910⁻¹⁰10⁻⁸10⁻⁶10⁻⁴10⁻²10³10³.⁵10⁴forcing term ηinner iterationsthe previous stepstarted at zerothe previous step, scaledthree curves within eight per cent of each otherand the unscaled guess is the worst of the three

A guess worth two per cent

The previous Newton step looks like a free guess at the next one, and it is worth nothing. Started from it unscaled, the inner solve costs 4 to 61 per cent more than starting from zero, because the guess is 15 to 209 times too large. Scaled by the ratio of the two residual norms it is the right size and halves the starting residual — which buys a constant handful of inner iterations, not a share, because conjugate gradients costs the logarithm of its tolerance.

sequence · Inexact newton
an eigenvalue repeated 2 times, in a 200-dimensional problemblock of 1never returns them allblock of 232 products, 16 stepsblock of 351 products, 17 stepsblock of 460 products, 15 stepsblock of 570 products, 14 stepsblock of 684 products, 14 stepsthe narrowest that workscheapest width2its products with A32the widest drawn84narrower than the multiplicity is not slowit is absent

How wide the block should be

A block narrower than the multiplicity does not converge slowly — it never returns the missing copy at all. Above the multiplicity every extra column buys iterations at about ten products with A each. And the mechanism that is supposed to make the choice unimportant never fires from a random start.

spectra · Invariant subspace
10²10⁴10⁶10⁸10¹⁰0285684112140168196224condition numberstepsnormal equationsbidiagonalisationone sequence, two costssteps at κ = 10², both16at κ = 10⁶, ratio1.1at κ = 10¹⁰, ratio1.9the same iterates in the algebraand twice the work at κ = 10¹⁰

One sequence and two recurrences

CGLS and LSQR compute the same iterates — the minimiser over a space is unique, so there is nothing to choose between them in the algebra. At κ = 10⁶ they cost 42 steps and 47. At κ = 10¹⁰ they cost 110 and 209, across four seeds, and the quantity that separates them is the orthogonality of a basis neither of them keeps.

iterative · Krylov
0481216202410⁻¹⁷10⁻¹⁴10⁻¹¹10⁻⁸10⁻⁵10⁻²10¹matrix–vector products, mrelative error in eᴬbforming e^A: 3·10⁻¹⁶crosses at m = 18the vector, not the matrixsteps to the dense answer18dimension100Krylov megaflops0.36dense megaflops2the exponential that is computedis 18×18

The vector was what was wanted

Nobody who computes a matrix exponential wants the matrix. They want eᴬᵗb — one vector, the state of a system at a later time. Twenty matrix–vector products get it to sixteen digits on a hundred-by-hundred problem, without ever forming a hundred-by-hundred exponential, and the exponential that does get computed is twenty by twenty.

spectra · Matrix function

An operator with no entries

At the sizes where linear algebra is expensive the matrix does not exist. What exists is a subroutine that returns Av. Every Krylov method survives that unchanged; every algorithm that reads an entry disappears. And the derivative such a code computes is accurate to ten digits instead of sixteen, which turns out to cost nothing at all.

iterative · Matrix-free

The zero that means it is finished

Every Krylov method ends by dividing by a number the previous step produced, and when that number is zero the recurrence stops. In Arnoldi the stop is the answer — the subspace has closed, the solution is inside it, and the residual is at the unit roundoff. The literature calls it a lucky breakdown, and the adjective is doing real work.

iterative · Breakdown

The same zero, and nothing was found

Change the recurrence by two lines and the divisor stops being a norm. It becomes an inner product of two vectors from two different sequences, and an inner product of two different vectors is zero on a whole hyperplane — with neither vector anywhere near zero, nothing invariant, and nothing converged. The arithmetic event is identical and the meaning is opposite.

iterative · Breakdown

The division that cannot be done

Conjugate gradients divides by pᵀAp at every step, and on a matrix that is not positive definite that number can be zero or negative. The guard against it has been here from the first essay and described it as a failure. In the method that made conjugate gradients famous it is the single most valuable object the iteration can produce, and it costs six matrix–vector products.

iterative · Breakdown

An eigenvalue that arrives twice

A matrix with forty distinct eigenvalues, handed to Lanczos for eighty steps, returns twenty-five extra copies of thirteen of them — the largest arriving five times. Every copy is accurate to 1.9·10⁻⁸ relative. No arithmetic error was made, nothing overflowed, and a caller counting eigenvalues gets the wrong multiplicity from a computation in which no individual number is wrong.

spectra · Lanczos

The residual the method reports

Conjugate gradients prints a relative residual of 6.9·10⁻²¹. The unit roundoff is 1.1·10⁻¹⁶, so that is not a small residual and not a large one — it is not a residual. The vector the method is holding at that step has ‖b − Ax‖/‖b‖ = 5.1·10⁻¹⁰, and nothing in the run says so.

iterative · Residual gap

The number that is re-derived

GMRES prints a residual it never computes from its answer either. On the matrix that sends a conjugate gradient recurrence 7.3·10¹⁰ wrong, and on two others chosen to be worse, its number is never more than a factor of 2.86 out — while the basis it is computed from has lost orthogonality entirely. The disease is not iterative methods, and it is not floating point.

iterative · Residual gap

An iterate that must be made smaller

Applying a Kronecker-sum operator to a low-rank iterate multiplies its ranks by d and adding two of them adds their ranks, so a solver in a compressed format cannot keep what it produces. Every step is followed by a truncation — and whether that truncation is a floor on the residual depends on the right-hand side rather than on the truncation.

iterative · Low-rank iteration

A Krylov space for a problem that is not linear

A quadratic eigenvalue problem has no matrix to build a Krylov space out of. The recurrence that builds one anyway stores half as many numbers, returns twice as many Ritz values — and stops being a basis at twenty vectors while the answer it gives keeps improving.

iterative · Krylov

A run that is over at step five

A conjugate gradient whose every iterate is cut to a rank budget reaches the floor that budget allows at step 5, 36, 42 or 59, and then does nothing for the rest of the run. Four times the iterations move the floor by a factor of 1.8, and past the answer's own rank they move it the wrong way.

iterative · Low-rank iteration

The same budget, spent five ways

A restarted method has one budget — products with A — and two ways to spend it, in many short cycles or a few long ones. At about a hundred and forty products the answer is the same to a factor of seven whichever split is chosen, and the residual bound the method reports spans ten orders of magnitude across the same five runs.

spectra · Lanczos

A proof that does not ask how large the matrix is

Proving a Hessian indefinite costs three matrix–vector products when the negative eigenvalue is 3 and nine to eleven when it is a thousandth, and that pair of numbers barely moves across a fourfold range in n. The factorisation that settles the same question costs a third of n³, which grows by a factor of sixty-four over the same range.

iterative · Breakdown

A walk needs a length

The gap between the two residuals grows as the square root of something, and a square root needs a length. Two quantities are candidates — how far the iterates travelled and how many steps were taken — and only a second sweep separates them. Across a fourfold change in size the iteration count goes from 39 to 96 and the gap goes from 5.04·10⁻¹⁵ to 5.33·10⁻¹⁵.

iterative · Residual gap

The answer that arrives when the space runs out

A second-order Krylov recurrence holds vectors of length n for a problem with 2n eigenvalues, so it is exact at n steps where the linearised route needs 2n. The machine-precision reading at forty-four vectors on a chain of forty is that exhaustion rather than convergence, and it arrives through a basis whose ‖QᵀQ − I‖ is above one.

iterative · Krylov

The certificate that arrives soonest is worth least

The more negative a Hessian's smallest eigenvalue, the sooner conjugate gradients meets a direction of negative curvature — and the less of the exact trust-region decrease that direction turns out to be worth. At λₘᵢₙ = −10 the step arrives after two products and gets 39.6 per cent; at −10⁻³ the same two products get 89.8, and the whole sweep costs eight.

iterative · Trust-region

Named alongside it

The objects these essays reach for when they reach for this one.

Conjugate gradientsCondition numberOrthogonalityResidualRitz valuesStopping criterionFlop countReorthogonalisationInvariant subspaceLoss of orthogonalityArnoldiFilter factors

All concepts