Concept

Tikhonov regularisation — where it appears

Adding a penalty on the size of the answer to a least-squares problem, which weights each singular component by σ²/(σ² + λ²). The parameter λ is chosen from the data by a rule rather than derived, and every published rule is a heuristic scored against a truth that usually does not exist.

Named by 31 essays across 2 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
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
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
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
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
081624324048566410⁻¹⁷10⁻¹⁴10⁻¹¹10⁻⁸10⁻⁵10⁻²10¹index kmagnitudethe floor: k = 43best truncation: k = 31σₖ|uₖᵀb| exact|uₖᵀb| with noise|uₖᵀe|, smoothedcorrelation ρ = 0.9noise energy, first 28 directions0.98tilt of the noise floor, decades1.1the crossing43the best truncation31the floor tilts, and is still a floorρ = 0.9

Noise that spares the answer and fools the rules

Make each noise sample remember the last one, keep its size fixed, and the best answer available gets slightly better — 0.1056 to 0.1010 — because slow noise hides in the directions where dividing by σ costs nothing. The Picard crossing still lands two dozen indices past the best truncation. What breaks is the rules. Generalised cross-validation more than doubles the best error on 14 draws of 48 instead of 3, the discrepancy principle's typical cost triples, and the two miss in opposite directions. Whitening by the covariance takes GCV back to 3.

regularisation · Regularisation
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
10⁻⁸10⁻⁷10⁻⁶10⁻⁵10⁻⁴10⁻³10⁻²10⁻¹110⁻⁴10⁻³10⁻²10⁻¹110¹10²10³10⁴λnoise part ÷ signal part, in ‖x‖where every corner sitsthe corner, ×2.29the oraclethe corner reads ‖x‖noise share at the corner0.23noise share at the oracle0.026corner ÷ oracle, this draw2.3corner ÷ oracle, 60-draw median1.5noise share: ‖L·(noise part)‖ ÷ ‖L·(signal part)‖the corner reads that share

The corner reads the norm it is drawn in

The L-curve was the costliest rule this field scored, and the cost was not the rule's. On the same sixty draws, with the same best achievable error, the corner of ‖x‖ against the residual costs 1.53 times the oracle and the corner of ‖L₁x‖ costs 1.003. Across five signals and three penalties the corner lands wherever amplified noise is between a tenth and a fifth of the norm being plotted, and it finds the oracle only when the oracle happens to sit there — twenty-nine times too costly on a smooth signal under ‖x‖, within half a per cent on four spikes.

regularisation · Parameter choice
unpreconditionedbest step20best error0.14Tikhonov's best0.14α = 0.001best step1best error0.14eigenvalues sent near one22051015202530354010⁻¹110¹steprelative errorTikhonov's best: 0.1405plain CGLSpreconditioneda better preconditionerarrives at the noise sooner

A preconditioner that arrives past the answer

On a system that is solved to convergence a preconditioner changes how fast the answer arrives and not what it is. On a problem regularised by stopping it changes where every step lands. Conjugate gradients preconditioned by AᵀA + αI reaches its best answer in one step at α = 10⁻³, and at α = 10⁻⁶ its best answer is its first step, with an error of 1.35 against the unpreconditioned run's 0.1426 — while the count of eigenvalues it has clustered at one rises from 22 to 32.

combination · Iterative regularisation
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
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
unpreconditionedbest step20rule stops at7its error ÷ best1steps within 10%23truncated at τ = 0.032best step8rule stops at4its error ÷ best1steps within 10%6010203040506010⁻¹110¹steprelative errorrule stops: 7rule stops: 4plain CGLSpreconditionedbars: the steps within 10% of each run's bestthe rule reads the residual, not the window

A stopping rule that follows the run it is given

A preconditioner that reaches the answer four times sooner leaves four steps within 10% of its best instead of sixteen, and a rule that stops by the residual ought to miss so narrow a window more often. Over forty draws of the noise it misses it less: the discrepancy principle stops at 1.030 times the preconditioned run's best against 1.073 times the plain run's. And past the edge it stops within a factor of 1.7 of a run whose own best is 5.5 times Tikhonov's — faithful to a run that has already failed.

combination · Iterative regularisation
unpreconditionedbest step20best error0.14Tikhonov's best0.14cosine, truncatedbest step5best error0.14directions preconditioned25010203040506010⁻¹110¹steprelative errorTikhonov's best: 0.1405plain CGLSpreconditionedthe blur approximated by a cosine transformtruncation leaves it alone

What a cheap preconditioner has to leave alone

A blur approximated by a matrix the cosine transform diagonalises agrees with the operator everywhere but its first and last seven rows. Made invertible by a shift, as the exact preconditioner was, it never reaches the unpreconditioned run's floor — at α = 10⁻³ its best iterate is 0.749 against 0.143. Made invertible by leaving every eigenvalue below τ alone, it reaches 0.141 in five steps instead of twenty, and the smallest τ that keeps the floor sits at a third to a half of the Tikhonov oracle's λ at three noise levels.

combination · Iterative regularisation
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
unpreconditionedbest step20best error0.14edge τ0.01cutoffs at τ = 0.5λdiscrepancy0.032GCV0.016L-curve0.00410⁻⁴10⁻³10⁻²10⁻¹10⁻¹110¹truncation τbest relative errorunpreconditioned floor: 0.1426discrepancy: 0.144GCV: 0.142L-curve: 0.234the cutoff read as 0.5 × each rule's own λthe plateau is wide and the cliff is steep

The rule that is wrong in the right direction

The preconditioner's cutoff is not a new parameter. It is the regularisation parameter this field already knows how to choose, halved — and the rule criticised for choosing λ a factor of two or three too large is the one whose cutoff keeps the floor on every draw, where the rule that chooses λ to within 3% has a worst draw two hundred times off it.

combination · Iterative regularisation
where each λ sitsoracle's share0.0037corner's share0.1corner's cost25and what a target costsbest fixed target0.0035its cost1estimate ÷ truth at the oracle0.9210⁻⁸10⁻⁶10⁻⁴10⁻²110⁻³10⁻²10⁻¹110¹10²10³10⁴λnoise part ÷ signal parttarget 0.0035oracle λ: share 0.0037corner: share 0.102solid: the share on this draw · dashed: the share a rule can estimatethe rule is a level set of the dashed curve

A rule that has to be told how good its answer will be

The L-curve's corner reads a noise share of 0.10 to 0.21 across fifteen pairings of signal and penalty, and the share the best λ sits at runs from 0.0037 to 0.43 — a factor of a hundred and fifteen. A rule aimed at the right share is within a few per cent of the oracle on every one of them. The right share is about a third to four-fifths of the relative error that λ will achieve, which is the number the answer was wanted for.

regularisation · Parameter choice
unpreconditioneddimension a step1.2best iterate carries24at step20τ = 10⁻³, past the edgedimension a step3.6best iterate carries33its error0.4837111519232731353910⁻¹10⁻⁰.⁵1effective dimensionrelative errorthe arc: dimension 8.0 to 24.0none10⁻⁰.⁵10⁻¹10⁻¹.⁵10⁻²10⁻².⁵10⁻³large dots: each run's own best iteratethe answer sits at one effective dimension

The parameter neither knob is

A preconditioned run has a cutoff and a step count, and neither is the regularisation parameter. The parameter is the effective dimension of the iterate: every cutoff that works puts its own best at 23.7 to 24.3 of it, where the unpreconditioned run's best sits at 23.2, and what the cutoff buys is the rate — 1.27 of it a step with no preconditioner and 3.53 with one. The edge is where a single stride is longer than the distance left.

combination · Iterative regularisation
what each reachesbest pair0.0034‖x‖ alone0.0048‖L₁x‖ alone0.0035λ₁, on ‖x‖λ₂, on ‖L₁x‖ — log₁₀0-6-5-4-3-2-100-6-5-4-3-2-10outlined row and column: one penalty switched offred cell: the best pair, 0.00343darker is worse, on a logarithmic scalethe optimum sits on or beside an edge

A second penalty is not a second parameter

Penalise ‖x‖ and ‖L₁x‖ at once and there are two λ to choose. Over ten draws on five signals the best pair beats the better single penalty by between 0.00% and 3.1%, and one of its two parameters is exactly zero on 30 to 70 per cent of draws. Choosing the wrong one of the two costs up to 54%. The surface is a choice between two curves with a knob nobody needs.

regularisation · Parameter choice
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
where each run leaves the arcnarrow blur, directions33the collection's blur, directions22wide blur, directions15the answer's own dimensionnarrow blur32the collection's blur23wide blur1501020300123456directions the preconditioner divideseffective dimension a stepnarrow blurthe collection's blurwide blurrings: the first cutoff past which most of the run misses the arcthey sit at three different strides

A count that marks the edge and not the pace

The number of directions a truncated preconditioner divides is counted for free when it is built, and it was proposed as a stand-in for the stride it buys. On three blurs it is not one — the runs leave the answer at strides of 5.71, 3.44 and 2.41. What the count does predict is the edge: on all three blurs, at two noise levels, a run stops landing on the answer's path within five per cent of the point where the count reaches the answer's own effective dimension. And the halved cutoff rule, measured on one blur, crosses that line on the narrowest.

combination · Iterative regularisation
falls through a half, % of the answerfast, truncated96exact, shifted105fast, shifted66the answer's own dimensionblur 2.5 points wide2300.250.50.7511.251.500.20.40.60.81the preconditioner's own dimension ÷ the answer'sshare of the run on the arcfast, truncatedexact, shiftedfast, shifteddashed: as much dimension as the answer carrieshorizontal: half the run on the arc

The shift had an edge, and the approximation moved it

A fast-transform preconditioner made invertible by a shift was recorded as never reaching the unpreconditioned floor, and predicted to sit off the answer's path at every shift. At a large shift it sits on the path and reaches the floor to a tenth of a per cent. It has an edge like the truncated one — but on the exact operator that edge is where the shift's own effective dimension reaches the answer's, 1.02 to 1.05 of it on six problems, and on the fast approximation it arrives at 0.49 to 0.77. The difference is sixteen samples at the ends of the signal, where the approximation is wrong and a shift divides the error by α.

combination · Iterative 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

One arc, and what each filter pays to be on it

Conjugate gradients and Tikhonov stop at the same effective dimension, and that could have meant two curves crossing once or one curve. It is one curve over a stretch — on six problems, Tikhonov and truncation reach the iteration's error at the iteration's dimension to within 7.2 per cent from 0.7 of the answer to its top — and the two separate on either side. But the curve is shared by a trade, not by an identical answer: at the same dimension Tikhonov carries 22 to 42 per cent more noise than the iteration and up to five per cent less bias.

combination · Iterative regularisation

The data count their dimensions, not the step's

Every grid in the deconvolution essays was chosen with the answer in hand, and so was every λ. From the data alone, the discrepancy principle's worst draw is within 16 per cent of the oracle on every grid from 16 points to 96; generalised cross-validation is better on the median draw and, on grids of thirty points and more, has draws thousands of times worse. And the data can say how many dimensions they carry — about 20, 25 and 29 at three noise levels, one number once the grid exceeds it — but not how many more the step needs: the grid that count chooses is 14 to 19 per cent worse than forty points at the lower two.

regularisation · Regularisation

The overshoot was the lead

Past its best, conjugate gradients' error rose more slowly than Tikhonov's at the same effective dimension — on the narrow blur at 0.1% noise Tikhonov was 2.35 times worse at 1.3 of the answer — and the reading was that the iteration spends its dimension where the data has content. Count admitted directions instead of summing factors, so that a factor of 1.81 counts once, and the lead is gone: 0.96 on that problem, and within 0.17 of one on all six from 1.1 to 1.3. Tikhonov now carries less noise than the iteration there. Below half the answer, where the three filters also disagree, the count changes nothing.

combination · Iterative regularisation

The minimum on the right

Generalised cross-validation's worst draws on a fine grid were all one mistake: a second dip in its function at λ near zero, deeper than the real minimum. The proposed repair was a residual threshold, one number, refusing any λ whose residual falls too far below the real minimum's. Measured over 528 draws, a threshold of one half still lets two hundredfold misses through; only the extreme value, which is no threshold at all but the rule 'take the rightmost local minimum', removes all eleven. It costs nothing on the coarse grids where the dip is the right answer, and on the collection's own problem over five thousand draws it turns 243 tenfold misses into 17.

regularisation · Regularisation

A tail from Tikhonov and a corner from truncation

Below half the answer, conjugate gradients beat Tikhonov at a matched count of directions by up to 69 per cent, and the proposed measurement was the sharpness p of a roll-off between Tikhonov and truncation that matches the iteration there. Any p from 2 to 4 closes the gap to a tenth; truncation, the family's limit, reopens it to a third. But no p describes the iteration. Its filter has Tikhonov's slope exactly in its tail and a local sharpness of 2 to 5 on its shoulder, and a single fitted p is a compromise that drifts from 2.3 at the first step to 1.7 at the answer.

combination · Iterative regularisation

More samples take the floor and leave the dip

GCV's catastrophic misses on fine grids were blamed on squareness: on an n × n system the residual and n − t both reach zero as λ does, and their ratio can dip there. With more samples than unknowns neither reaches zero, and the prediction was that the dip would be gone by construction. Half of it is. The minimum at the floor of the scale, 21 draws in 240 on square systems, is gone at every ratio. The interior dip is not — 28, 20, 13, 8 and 4 draws at one to four samples per unknown — and at four per unknown one draw still misses the oracle by 877 times.

regularisation · Regularisation

A third penalty on a flat floor

Two penalties at once were worth three per cent at most, and the explanation offered was that one penalty already does the work — which predicts that a third buys less still and that the best choice sits on a face of the parameter cube. The third buys a median of exactly nothing on all five signals and 0.66% on its best draw. But on twelve draws of forty the best triple does use all three, and on every one of them the nearest point with a penalty switched off is within that same 0.66%. The minimum is not on a face; it is on a floor so flat that where it lands is noise. Choosing the right single penalty is worth a factor of two.

regularisation · Parameter choice

Named alongside it

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

Filter factorsParameter choiceDiscrepancy principleIterative regularisationRegularisationConjugate gradientsDeconvolutionSemi-convergenceIll-posed problemGeneralised cross-validationL-curveEffective dimension

All concepts