Concept

Ill-posed problem — where it appears

A problem whose singular values decay to nothing with no gap, so the data does not determine the answer and something outside it must choose. It is the situation regularisation exists for, and the reason a solver's job stops being arithmetic and becomes a choice about how much of the data to believe.

Named by 26 essays across 5 fields — each of them below, with the objects they name alongside it.

081624324048566400.250.50.751index kfilter factor fₖno regularisation: fₖ = 1truncationTikhonovthe same sum, three weightsTikhonov, relative error0.11truncation, relative error0.11no filter at all5.5·10⁸both filters are one expression with a different weightfₖ = 1 is the catastrophe

When the answer is a choice

A backward-stable least-squares solve of this problem returns an answer whose relative error is 5.5·10⁸. Nothing went wrong. The singular values decay exponentially with no gap anywhere in them, the data does not determine the answer, and something outside the data has to choose — which is the computation rather than a preliminary to it.

regularisation · Regularisation
00.250.50.75110⁻¹110¹10²10³fraction of the method's own rangerelative errorfloor 0.141truncation KTikhonov λCGLS steprandomised rankfour methods, one floortruncation K0.14Tikhonov λ0.14CGLS step0.14randomised rank0.14four knobs from four fieldsand one obstruction underneath them

Four knobs and one floor

A truncation, a Tikhonov parameter, a step count and a randomised rank, on one problem with an answer that is known. Their best errors are 0.1445, 0.1406, 0.1426 and 0.1449 — a spread of 3% across four methods that share no arithmetic.

combination · Parameter choice
081624324048566410⁻¹⁷10⁻¹⁴10⁻¹¹10⁻⁸10⁻⁵10⁻²10¹index kmagnitudethe floor: k = 32best truncation: k = 28σₖ|uₖᵀb| exact|uₖᵀb| with noisetwo different indicesthe crossing, from the data alone32the truncation that is actually best28relative error there0.11the exact coefficients never flattenthe noisy ones stop at ‖e‖/√n

Where the answer stops being in the data

The Picard condition finds the index where a noisy right-hand side stops carrying signal, from the data alone, with no knowledge of the answer. It lands at 32 where the truncation that actually minimises the error is 28 — and at 45 where the best is 38. It overshoots at every stop from 10% noise to 0.0001%, and it overshoots for a reason. The best truncation walks up the spectrum in a straight line, six or seven indices a decade; the crossing climbs in jumps of 11, 0, 8, 5 and 1.

regularisation · Regularisation
110¹10²10³10⁻³10⁻²10⁻¹110¹10²10³n‖Aₙ − A‖ and the largest term's norm‖Aₙ − A‖the larger of its two termsan infimum that is not attainedn1024‖Aₙ − A‖0.0017largest term1024their product1.7√31.7the distance goes to zeroand nothing reaches it

A nearest point that is not there

Eckart and Young guarantee that a matrix has a best rank-k approximation and that the truncated SVD is it. For three indices the guarantee is false in the strongest available way — there are tensors whose distance to the rank-two set is zero and which no rank-two tensor equals.

tensor · Tensor rank
10⁻²10⁻¹110¹10²10³10⁴‖Ax − b‖‖x‖the oraclediscrepancyL-curvegeneralisedscored against a truth none hasoracle, relative error0.11discrepancy principle, as a multiple1.1L-curve corner, as a multiple2.3generalised cross-validation, as a multiple1the oracle needs the exact answer and is not a methodit is the reference the others are scored on

Choosing without knowing

Three published rules for choosing a regularisation parameter, scored against an oracle that requires the exact answer and is therefore not a method. Generalised cross-validation lands on the oracle's λ exactly; the discrepancy principle costs 6%; the L-curve costs 129%. And told a noise level ten times too small, the discrepancy principle's error goes from 0.112 to 10,449.

regularisation · Parameter choice
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
0481216202428321rank keptrelative errormedianthe answer movesspread at rank 81.8spread at rank 241best median error0.14widest where the method is worstand the bound does not say so

An answer that changes with the seed

A randomised rank-k solve is a truncation computed in a random subspace, and it reaches the same floor as the deterministic ones. What it does not do is return the same answer twice — a factor of 1.84 across four seeds at rank 8, and 1.02 at the rank where the method is best.

combination · Randomised
192123252729313335373941434500.10.20.30.40.50.60.7position jweight on the true signal at jthe blur's row, 5.89 widethe kernel, 2.82 widecomputed from A and λ alonekernel width at half height2.8components kept, Σfₖ28width × Σfₖ / n1.2deepest negative lobe-0.075no data and no truth went into this curvethe answer is the truth seen through it

A second blur, narrower than the first

A regularised answer is not the truth with the noise taken out. It is the truth seen through a second blur, V F Vᵀ, which depends on the operator and λ and on nothing that was measured. At the best λ for 0.1% noise its rows are 2.82 points wide against the instrument's 5.89, they dip to −0.075 on either side, and their width times the number of components kept stays between 1.10n and 1.27n across seven decades of λ. Two spikes four points apart come back as two; three apart, as one.

regularisation · Regularisation
1591317212529333710⁻¹110¹10²bidiagonalisation stepsrelative errorleast without: 20no penaltypenalty insidewhat stopping is worthbest without a penalty0.14and at step 40161best with one0.14and at step 400.14the same floor, reached twiceand only one run stays on it

The step that stops mattering

Regularise the problem the iteration has built rather than the problem it was given, and the error curve stops turning. The unregularised run ends 1,127 times above its own best; the same run with a penalty inside it ends 1.000000000003 times above.

combination · Iterative regularisation
27121722273237424710⁻²10⁻¹110¹subspace size kvalueno answer below 8GCV's traceλ from GCVλ from the residuala denominator that is not the problem'strace at k = 41trace at k = 4827λ range across the run2.2subspace before an answer8the divisor moves by twenty-sevenand the answer does not move

A parameter chosen on a smaller problem

Inside a hybrid method the regularisation parameter is chosen on a 25×24 problem rather than a 64×64 one. The rule that reads a residual transfers exactly; the rule that reads a trace is biased by exactly two grid steps at twenty-four steps and one at forty, at every noise level from 10% to 0.1%.

combination · Parameter choice
110¹10²10³10⁴10⁵10⁶error ÷ the oracle's, on the same drawten times the oracleover tendiscrepancy principletold ‖e‖0 of 1000unbiased risktold σ²15 of 1000cross-validationtold nothing57 of 1000quasi-optimalitytold nothing0 of 1000L-curvetold nothing0 of 1000blue: median · bar to the 90th percentile · line to the 99th · red: worstthe counts are the tail

One draw in twenty

Sixteen draws gave generalised cross-validation a worst case of 12%. A thousand draws at each of five noise levels give it a second answer on four to six in every hundred, ten to seven million times worse than the oracle, while its median stays among the best of five rules. The quasi-optimality criterion, told nothing either, never costs more than 1.41 in five thousand draws. The share settles by a thousand draws, and letting the search look further down more than triples it.

regularisation · Parameter choice
110¹10²10³10⁴10⁻¹⁴10⁻¹¹10⁻⁸10⁻⁵10⁻²10¹sweeprelative error, and the largest term's sizerising: the swamp's largest rank-one termfalling, slowly: its errorfalling, once: a fit with an answera plateau with a rising floorswamp error0.0014swamp term10term growth2.2benign error9.7·10⁻¹⁵benign term growth1the error alone cannot tellthe size of the terms can

An iteration that walks out of the set

Every sweep of alternating least squares is the exact minimiser of its own subproblem, so the objective can only fall. What it cannot do is converge, when the target's nearest rank-r point is not in the rank-r set — and a plateau at a small residual looks identical to slow convergence unless the size of the terms is plotted beside it.

tensor · Alternating least-squares
1216202428323640444810⁻¹110¹10²grid points nrelative error against the continuous signalbest truncation, 0.117best grid: n = 26best K = 28with noisenoise-freeno λ anywhere in this curvebest grid, error0.13best truncation, error0.12κ on the best grid61grid where it has doubled34every point is an unregularised solvethe grid chose the truncation

The grid was the first filter

A continuous deconvolution discretised on n points and solved with no regularisation at all is not unregularised. Its error against the continuous signal is least at 24, 26 and 34 points for noise of 1%, 0.1% and 0.01% per sample — beside best truncations of 24, 28 and 32 components on a 64-point grid — and within 4 to 16 per cent of their error. The grid's own filter factors sum to n exactly and fall through a half at k = n. Choosing the grid was choosing a truncation, before anybody chose a λ.

regularisation · Regularisation
1110¹10²10³10⁴10⁵the noise level it is told ÷ the true oneerror ÷ the oracle's0.40.50.71.523twice the oraclecliff, ρ = 0.69told the truthworst of 400middle 80%median draweach draw against its own oraclecliff, median ρ0.69told the truth, median1.1told a half, median3201shaded: the middle 80% of drawsbelow the cliff the error has doubled

Thirty-two coefficients instead of a noise level

The discrepancy principle has to be told the noise, and told too little it does not degrade — it falls off a cliff, at 0.80 of the truth when the noise is 10% and at 0.58 when it is 0.001%, exactly where the understatement forces the filter past its best truncation. The missing number is in the data. The root mean square of the last thirty-two coefficients never sends the rule over the cliff at or below 1% noise in four hundred draws, where eight coefficients with the same median do so thirty-five times.

regularisation · Parameter choice
12345610⁻⁶10⁻⁵10⁻⁴10⁻³10⁻²10⁻¹digits of measurement accuracybest relative errorTikhonovtruncationCGLS stepslope 0.89, optimalslope 2/3, Tikhonov's limithow fast the error fallsTikhonov slope0.7truncation slope0.94CGLS step slope0.89optimal 2ν/(2ν + 1)0.89Tikhonov's limit0.67better data, smaller errorat a rate the method may not be able to keep

The method that cannot use a smooth answer

On the collection's own signal four regularisers reach the same floor to a few per cent. Score them instead against answers of increasing smoothness and one stops improving. Across six decades of noise Tikhonov's error falls with a fitted slope of 0.70 whether the answer is twice or four times as smooth, while truncation's rises to 0.94 — and at 10⁻⁶ noise Tikhonov's best is 38 times truncation's.

combination · Iterative regularisation
10141822263034384210⁻¹10⁻⁰.⁵110⁰.⁵grid points nrelative error against the continuous signalsampled kernel, hatsintegrated, hatsintegrated, spline0.1% noise, no λsampled kernel, hats, best0.13integrated, hats, best0.12integrated, spline, best0.12truncation of 64 points0.12same nodal unknowns, same databetter at representing the noise too

A better discretisation is a weaker filter

A coarse grid's error has two sources — how well the discrete operator approximates the integral, and how well the grid's function represents the answer — and the grid essay could not separate them. Changed one at a time they separate: integrating the kernel against the hat functions takes a fifth off the 12-point error, reading the answer as a cubic spline takes 15 per cent more, and both roughly double the condition number on every grid. At 0.1% noise the spline discretisation's unregularised solve on 26 points reaches the best truncation of a 64-point grid to 0.3%. At 1% it is worse than the crude grid.

regularisation · Regularisation
110¹10²10⁻¹²10⁻⁸10⁻⁴110⁴ncondition number, and residual reachedmarks above: κ of the step · dashes: its closed formmiddle: a fit from a random startbelow: the same fit started at the answera decomposition that is ill-conditionedκ at n = 1283.3·10⁴its closed form3.3·10⁴cosine of the terms1from a random start0.012from the answer1.4·10⁻¹⁰the answer existsand cannot be found

A tensor that cannot be decomposed

Every member of a certain sequence is exactly a sum of two rank-one terms, and both terms are written down in closed form. A three-hundred-sweep fit from a random start does not find them, and stalls at the same one per cent however far the sequence goes — while a fit started at the answer loses digits exactly as 2n² says it should.

error · Conditioning
122028364452606876849210010⁻¹110¹grid points nrelative error against the continuous signalbest with no λ: n = 26no λeach grid's best λTikhonov, 64 points0.1% noise, 16 drawsbest grid with no λ26its error0.1396 points with its λ0.12best λ on fine grids0.0056each point is a median over the same drawsλ belongs to the problem, not to the grid

Where the grid hands over to λ

An unregularised solve on a coarse grid comes within a tenth of the best Tikhonov answer on a fine one, and the pair of a grid and a λ was left unmeasured. Measured, the two do not trade. On every grid up to the best unregularised one no λ helps at all. On every grid of 40 points and more the best λ is the same to within a quarter of a decade — 3.2·10⁻² at 1% noise per sample, 10⁻³ at 0.01% — and the 96-point grid with it beats the best coarse grid by 7, 9 and 13 per cent. The grids between the two, given their own λ, land between them.

regularisation · Regularisation
01210⁻¹⁶10⁻¹³10⁻¹⁰10⁻⁷10⁻⁴10⁻¹terms fitted1 − the largest cosine between two fitted terms3 — the tensor's rank45below this line, two terms are the same termthe residual says nothing3 terms, collinearity0.524 terms, collinearity15 terms, collinearity13 terms, error2.1·10⁻¹⁴4 terms, error10⁻¹³4 terms, sweeps76one term too manyand two of them cancel

One term too many

A tensor built from three rank-one terms, fitted with four, reaches a relative error of 1.0·10⁻¹³ in seventy-six sweeps. Two of the four terms come back with a cosine of 0.99927 between them, subtracting each other, and the smallest weight is a fifth of the largest rather than a rounding. Nothing in the residual says any of it.

tensor · Alternating least-squares
10⁻¹⁰10⁻⁸10⁻⁶10⁻⁴10⁻²10⁻¹⁰10⁻⁸10⁻⁶10⁻⁴10⁻²1λ, the ridge on every subproblemerror, and the largest rank-one termthe terms a swamp was growingthe error it costs a fit that has an answerthe error it costs one that has nota bound, at its own priceterms, no ridge6.7terms, heaviest1.2sweeps, no ridge3000sweeps, heaviest69benign error at λ0.089terms still the right terms1the plateau is boundedand the answer is biased by exactly λ

The repair that costs exactly itself

A ridge on every subproblem is the standard cure for a swamp and it works — the terms stop at 1.61 instead of climbing past 6.7, and a run that never finished finishes in 798 sweeps. On a tensor that does have an answer the error it costs is the ridge itself, to within a factor of two, at every setting from 10⁻¹⁰ to 10⁻¹. And the terms it returns are still the right terms.

tensor · Alternating least-squares
01210¹10²10³the three kinds of runsweep at which the test firesno answer to reachbadly conditionedordinaryhollow: never fired, drawn where the run endedwhat it fires on, and whenthreshold0.25boundary, fired6boundary, median sweep22collinear, fired6collinear, median sweep759ordinary, fired0the value does not separate themand the sweep does

The test that is a deadline

The quantity that separates a boundary from slow convergence fires on every run that has nothing to reach, at sweep 21 of four thousand, and on no ordinary fit. It also fires on every badly conditioned fit that does converge — at sweep 758. No threshold between 0.10 and 0.40 separates those two, and the sweep it fires at separates them by a factor of thirty-six.

tensor · Alternating least-squares
1220283644526068768492100101214161820grid points nrelative error against the continuous signal, per centsampled kernel, hatsintegrated, hatsintegrated, splinespline, no λ1% noise per sample, median of sixteen drawssampled kernel, hats: 16 points0.18integrated, hats: 16 points0.16integrated, spline: 16 points0.16sampled kernel, hats: 96 points0.14integrated, hats: 96 points0.14integrated, spline: 96 points0.14spline, no λ, its best grid0.15dotted: the 96-point answerevery point is a grid with its own λ

The grid on which the discretisation stops mattering

Without regularisation, integrating the blur's kernel against cubic splines beat sampling it at 0.1% noise and lost to it at 1%. Give each discretisation its own best λ on every grid and the difference shrinks to nothing where grids are fine — 0.17, 0.28 and 0.20 per cent apart on 96 points at the three noise levels, with every discretisation choosing the same λ — and stays at 17 to 18 per cent on 16 points. The choice between them is a choice of how coarse a grid can be: at 0.1% noise the integrated discretisations reach the fine-grid answer on 26 points and the sampled one needs 40.

regularisation · Regularisation
on 96 pointsno breakpoint0.12two jumps0.05doubled knots0.13one box0.013the box on fewer points16 points0.0526 points0.02248 points0.00710203040506070809010010⁻²10⁻¹.⁵10⁻¹grid pointsrelative error, best λno breakpointtwo jumpsdoubled knotsone boxevery reading spends the same n unknownsand is given its own best λ on each grid

A corner the penalty can afford

Every smooth reading of the deconvolution's grid needed about forty points and then stopped improving, and the step was the suspect. Give the step one coefficient of its own and forty-eight points reach an error of 0.0070 at 0.1% noise, against 0.118 for the best smooth reading on ninety-six — the step was most of the error. But the same step given two coefficients recovers half as well, and given a doubled node at each edge it recovers worse than no breakpoint at all, while representing the signal to 0.07%. What decides is what the penalty is charged for the corner, and whether the data can say where it is.

regularisation · Regularisation
at rank threelowest across tensors1highest across tensors1at rank fourlowest across tensors0.056highest across tensors0.25234500.20.40.60.81rank fittedagreement among the starts that reach itsix planted tensorsdashed: the true rankeach line one tensor, fitted at ranks two to fivefive starting points at every rank

The rank a sweep can vouch for

To decide a tensor's rank, fit it at one rank after another and stop where something changes. Three things can be read off each rank's fits — the error, how nearly parallel the terms are, and whether fits from different starting points agree — and over six planted rank-three tensors at four noise levels they decide it 24, 12 and 15 times out of 24; the ratio of consecutive errors, which needs nothing, decides it 24 times. The error is right every time and has to be told the noise level. The collinearity is a coin toss. Agreement among starts is never wrong and, under noise, cannot decide on nine of the twenty-four — and the rank past the answer, where every decision is made, costs ten times the answer.

tensor · Alternating least-squares

A fit that has an answer and cannot stop

Fit a noisy rank-three tensor with four terms and the prediction was that the spare term would find real structure in the noise and stop pairing off with the others. It does not: the largest cosine between two fitted terms has a median of 0.977 at 1% noise against 0.975 without noise. What changes is the solver. Without noise the four-term fit stops in fifty sweeps; with noise it never stops: its error keeps moving by a part in a million a sweep for six thousand sweeps, while on most tensors its terms stand still. The extra term takes exactly its share of the noise, and the error is settled by sweep 150.

tensor · Alternating least-squares

A problem with no answer

If two matrices share a null vector then det(A − λB) is identically zero and every λ is an eigenvalue, which means none of them is. Perturb such a pencil by a ten-billionth and a solver returns six numbers with residuals below 10⁻⁹. Change the seed and it returns six different numbers, spread over forty-four, with residuals just as small.

spectra · Pencil

Named alongside it

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

Filter factorsRegularisationTikhonov regularisationTruncated SVDBorder-rankAlternating least-squaresCP decompositionDeconvolutionCondition numberParameter choiceSwampTensor rank

All concepts