Depth

Series — page 3

A field says what an essay is about. A series follows one idea essay by essay — from the question that introduces it to the one that assumes all the others.
-14-12-10-8-6-4-2010⁻¹⁷10⁻¹³10⁻⁹10⁻⁵10⁻¹10³10⁷10¹¹10¹⁵log₁₀ μ — the barrier parametercondition number, and relative errorκ₂, condensedκ₂, augmentederror, condensederror, augmentedagainst a BigInt answerκ₂ augmented, μ = 10⁻¹⁴3·10¹⁵its relative error10⁻¹⁵κ₂ condensed2.4·10¹⁶its relative error0.31the same step, written two waysand only one of them is solvable

Interior-point conditioning

  1. 1 A condition number sent to infinity
  2. 2 The active set before the digits
  3. 3 Two repairs for one symptom
  4. 4 A test with no tolerance in it
4 essays · constraint
10⁻⁹10⁻⁶10⁻³110³10⁶13212937κ · usignificand bitsκu = 1a bound was provedthe method refusedit never returns a wrong boundlargest κu with a proof0.45smallest κu without one0.89cases refused, of the grid13a refusal is not a wide bound — it is no bound at alland it is the only failure mode here

Interval

  1. 1 A bound that is proved
  2. 2 Proving the answer is in the box
  3. 3 Where the box is cut
  4. 4 Nine steps of pessimism
4 essays · arithmetic
λ = 105 timesλ = 9.55 timesλ = 95 timesλ = 8.55 timesλ = 2.952 timesλ = 2.92 timeseigenvalues that arrived more than once — the matrix has 40 distinct onesa spectrum with the wrong multiplicitiesextra copies, no reorthogonalisation25extra copies, full reorthogonalisation0worst relative error among the copies1.9·10⁻⁸steps taken of 80 asked for, full40no arithmetic error was madeevery one of these is right to eight digits

Lanczos

  1. 1 An eigenvalue that arrives twice
  2. 2 Restarting is a filter
  3. 3 Keeping the vectors, and losing the bound
  4. 4 The same budget, spent five ways
4 essays · spectra
0246810⁻¹⁷10⁻¹⁴10⁻¹¹10⁻⁸10⁻⁵10⁻²10¹log₁₀ γ, the change of unitsrelative errorforward errorη, the quadraticη, the linearisationagainst a closed formη(linearisation), worst7.6·10⁻¹³η(quadratic), worst1.2·10⁻⁴forward error, worst0.0013coefficient spread4.2·10¹⁵the solver is right at every stopabout a problem nobody asked

Linearisation backward error

  1. 1 A backward-stable answer to a problem nobody asked
  2. 2 Six routes to one spectrum
  3. 3 A perturbation that moves every coefficient
  4. 4 The number that moves when the problem does
4 essays · polynomial
10⁻¹²10⁻¹⁰10⁻⁸10⁻⁶10⁻⁴10⁻²10⁻¹⁴10⁻¹10¹²10²⁵10³⁸10⁵¹10⁶⁴δ, the gap between consecutive eigenvaluesrelative error in eᴬan answer with no correct digitsV f(Λ) V⁻¹scaling and squaring‖A(δ) − A₀‖exact eigenvalues throughoutκ(V) at the smallest δ3.3·10⁸²eigen route2.9·10⁶⁵scaling and squaring4.1·10⁻¹²distance to the limit7·10⁻¹²the eigenvalues are the diagonaland they are exact at every stop

Matrix function

  1. 1 A function of a matrix is not a function of its entries
  2. 2 The series that has to be squared back
  3. 3 The vector was what was wanted
  4. 4 The error the method already knows
4 essays · spectra
-2-101234-5-3-1135real partimaginary part4 insidea countable spectruminside the contour4drawn12existing∞worst branch residual1.6·10⁻¹⁵there is no last eigenvalueso the question has to change

Nonlinear eigenvalue

  1. 1 A problem with infinitely many eigenvalues
  2. 2 A ceiling with a knob on it
  3. 3 The conditioning that rises with the ceiling
  4. 4 Where a contour's budget should go
4 essays · polynomial
10¹10⁴10⁷10¹⁰10¹³10¹⁶10⁻¹⁷10⁻¹⁴10⁻¹¹10⁻⁸10⁻⁵10⁻²10¹κ(B)relative error, and asymmetryvia B⁻¹Avia Choleskyasymmetry of B⁻¹Au · κ(B)against a spectrum known exactlyslope, via B⁻¹A0.92slope, via Cholesky0.98worst ratio between them2.3asymmetry of B⁻¹A1.1the symmetry claim is trueand it is not about the accuracy

Pencil

  1. 1 Two matrices and one problem
  2. 2 An eigenvalue with no value
  3. 3 A problem with no answer
  4. 4 The largest gap is inside the null space
4 essays · spectra
-18-15-12-9-6-300log₁₀ ‖PKPᵀ − LDLᵀ‖ / ‖K‖share of orderingsbestworstexistence and stabilityfactorise1unregularised0.69worst growth6.4·10⁵growth × γ0.64the ordering is free to chooseand not free of consequence

Quasi-definite

  1. 1 The regularisation that legalises every order
  2. 2 The perturbation that does the work
  3. 3 A shift that certifies a saddle
  4. 4 The residual turns before the error doubles
4 essays · constraint
1234567891010⁻¹⁸10⁻¹⁵10⁻¹²10⁻⁹10⁻⁶10⁻³1indexsingular valuecutoff, σ₁ · 10⁻¹⁰numerical rank 10gap 8.2·10⁶an opiniontrue rank 410×10, built with 4 nonzero valuesrank is a decision

Rank

  1. 2 Rank is a decision
  2. 3 The cheap rank and what it cannot see
  3. 4 A rank that depends on the thread count
  4. 5 A good curve and a bad verdict
4 essays · spectra
10²10³10⁴10⁻¹⁷10⁻¹⁶10⁻¹⁵10⁻¹⁴10⁻¹³10⁻¹²steps taken‖RᵀR − AᵀA‖ ⁄ ‖AᵀA‖the bound, linear in the steps√k · uevery step safe, the chain notdrift after the run3.9·10⁻¹⁴the bound there3.3·10⁻¹³√k · u there6.1·10⁻¹⁵worst single amplification2.7worst leverage met0.69refreshes1backward stable onceand three thousand times is a different claim

Sequence stability

  1. 1 Stable once, and three thousand times
  2. 2 The repair the drift did not need
  3. 3 The answer the last window left
  4. 4 The step the two rows owe
4 essays · sequence
the smallest perturbation of any kind — 1.89·10⁻¹⁷the smallest Toeplitz one — 2.65·10⁻¹³both exact for the same x̂smallest of any kind1.9·10⁻¹⁷smallest Toeplitz one2.7·10⁻¹³the price of the constraint1.4·10⁴diagonal defect, unconstrained0.97an exact answer to a nearby problemof a kind nobody posed

Structured backward error

  1. 1 A nearby problem of the wrong kind
  2. 2 The condition number of the model
  3. 3 A perturbation that keeps the symmetry
  4. 4 The number that cannot rank them
4 essays · structure
545545545455554545545545545454555555454545545545545455554545545545rows and columns, in the order the points arrivedark: kept dense · light: two thin factors, rank printedthe strong partitionblocks112kept dense46largest rank5numbers stored2.7·10⁴relative compression error3.4·10⁻¹⁰the picture is decidedbefore a number is read

Admissibility

  1. 1 Which pairs are allowed to be small
  2. 2 The test that costs what it saves
  3. 3 The same matrix, numbered twice
3 essays · hierarchy
10¹10³10⁵10⁷10⁹10¹¹110¹10²10³10⁴condition number of the matrixgrowth factorbound 2^11partial pivotingCholeskyno pivot to gain fromCholesky growth, every κ1Cholesky interchanges0partial pivoting, worst9the bound, 2^112048both eliminations reach the same growthand only one of them had to swap to get there

Cholesky

  1. 1 A factorisation with nothing to pivot for
  2. 2 When symmetry is not enough
  3. 3 Where the multipliers go
3 essays · elimination
the estimator maximises this quantity over the columns it visitscolumn 1 ‹visited›12column 2 ‹the answer›114column 311.4column 411.4column 511.4column 611.4column 711.4column 811.4column 911.4column 1011.4column 1111.4column 1211.4estimate 12.0a walk that stopped earlythe estimate returned12the true 1-norm114columns visited1products with the matrix5the walk's own stopping test firedand every column it could see was smaller

Condition-estimation

  1. 1 An estimate that can be fooled
  2. 2 The tail a sample never reaches
  3. 3 Two columns see what one walk cannot
3 essays · error
110¹10²10³10⁴10⁵10⁶10⁷00.250.50.751amplification of the input perturbationfraction of directions at or belowκ = 10·10⁵worst found 7.6·10⁵6×6, 200 directionsmedian reaches 0.29 of κ

Conditioning

  1. 1 The condition number is an amplifier
  2. 2 A tensor that cannot be decomposed
  3. 3 The roots are not the coefficients
3 essays · error
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

Deliberate zero

  1. 1 The zero you are allowed to write
  2. 2 Deciding that a zero has arrived
  3. 3 A tolerance is priced by the problem
3 essays · error
10⁻¹⁶10⁻¹³10⁻¹⁰10⁻⁷10⁻⁴10⁻¹10⁻¹⁶10⁻¹³10⁻¹⁰10⁻⁷10⁻⁴10⁻¹size of the perturbation ‖δA‖how far the eigenvalues moveJordan block, ε^(1/8)symmetric, ≤ ‖δA‖rounding error alone moves it to 10⁻²six seeds per symmetric point; Jordan is closed formsymmetry beats precision

Eigen conditioning

  1. 1 Symmetry is worth more than precision
  2. 2 A condition number for one eigenvalue
  3. 3 The gap decides the eigenvector
3 essays · spectra
3456789101112110¹10²10³10⁴10⁵10⁶nwidest intermediate, in bitsrationals, not reducedfraction-free · rationals reduced · the answerone answer, three widthsn12answer40Hadamard bound52fraction-free40reduced rationals39unreduced1.4·10⁶the error is zero on every curvethe cost is the length of the numbers

Exact cost

  1. 1 An answer with no error in it
  2. 2 The answer is longer than the question
  3. 3 An exact answer to a measured problem
3 essays · exact
10⁻⁴10⁻³10⁻²10⁻¹110¹10²10³10⁻¹⁶10⁻¹³10⁻¹⁰10⁻⁷10⁻⁴10⁻¹relative change in the coefficients, along the worst directionrelative increase in the residualcoefficients doubled39% change, fit unmoved in the sixth digit308×: the third digit movesκ(A) = 3.6·10⁶. Exact arithmetic would pick one point on this floor. It would not raise it.24 points, degree 9, monomial basisthe data leaves them free

Fitting

  1. 3 The valley with no bottom
  2. 4 A basis built from the points
  3. 5 The degree that is safe to overshoot
3 essays · leastsquares
[½, 1)[1, 2)[2, 4)0.5124gap 0.125gap 0.25 — twice as wide8 values per octavespacing doubles at each power of two

Floating-point

  1. 1 What a float can hold
  2. 2 The other half of a format
  3. 3 Eight bits, and a format that breaks the rules
3 essays · arithmetic

All essays