Probabilistic Error Analysis for Inner Products

Probabilistic models are proposed for bounding the forward error in the numerically computed inner product (dot product, scalar product) between of two real $n$-vectors. We derive probabilistic perturbation bounds, as well as probabilistic roundoff error bounds for the sequential accumulation of the inner product. These bounds are non-asymptotic, explicit, and make minimal assumptions on perturbations and roundoffs. The perturbations are represented as independent, bounded, zero-mean random variables, and the probabilistic perturbation bound is based on Azuma's inequality. The roundoffs are also represented as bounded, zero-mean random variables. The first probabilistic bound assumes that the roundoffs are independent, while the second one does not. For the latter, we construct a Martingale that mirrors the sequential order of computations. Numerical experiments confirm that our bounds are more informative, often by several orders of magnitude, than traditional deterministic bounds -- even for small vector dimensions~$n$ and very stringent success probabilities. In particular the probabilistic roundoff error bounds are functions of $\sqrt{n}$ rather than~$n$, thus giving a quantitative confirmation of Wilkinson's intuition. The paper concludes with a critical assessment of the probabilistic approach.

[1]  D. Calvetti Roundoff error for floating point representation of real data , 1991 .

[2]  Françoise Chatelin,et al.  CESTAC, A TOOL FOR A STOCHASTIC ROUND-OFF ERROR ANALYSIS IN SCIENTIFIC COMPUTING , 1986 .

[3]  J. Richard Swenson,et al.  Tests of probabilistic models for propagation of roundoff errors , 1966, CACM.

[4]  William Kahan,et al.  The improbability of probabilistic error analyses for numerical computations , 1996 .

[5]  Roundoff error distribution in fixed point multiplication , 1980 .

[6]  James Hardy Wilkinson,et al.  Rounding errors in algebraic processes , 1964, IFIP Congress.

[7]  Mei Han An,et al.  accuracy and stability of numerical algorithms , 1991 .

[8]  Jesse L. Barlow,et al.  On roundoff error distributions in floating point and logarithmic arithmetic , 1985, Computing.

[9]  Ivo Babuska,et al.  On Roundoff Error Growth in Elliptic Problems , 2018, ACM Trans. Math. Softw..

[10]  M. Mitzenmacher,et al.  Probability and Computing: Chernoff Bounds , 2005 .

[11]  J. Neumann,et al.  Numerical inverting of matrices of high order , 1947 .

[12]  P. Henrici PROBLEMS OF STABILITY AND ERROR PROPAGATION IN THE NUMERICAL INTEGRATION OF ORDINARY DIFFERENTIAL EQUATIONS , 2010 .

[13]  Fan Chung Graham,et al.  Concentration Inequalities and Martingale Inequalities: A Survey , 2006, Internet Math..

[14]  Martti Tienari,et al.  A statistical model of roundoff error for varying length floating-point arithmetic , 1970 .

[15]  Nicholas J. Higham,et al.  A New Approach to Probabilistic Rounding Error Analysis , 2019, SIAM J. Sci. Comput..