Testing and Reconstruction of Lipschitz Functions with Applications to Data Privacy

We study local filters for two properties of functions f:{0,1} d → ℝ: the Lipschitz property and monotonicity. A local filter with additive error a is a randomized algorithm that is given black-box access to a function f and a query point x in the domain of f. Its output is a value F(x), such that (i) the reconstructed function F(x) satisfies the property (in our case, is Lipschitz or monotone) and (ii) if the input function f satisfies the property, then for every point x in the domain (with high constant probability) the reconstructed value F(x) differs from f(x) by at most a. Local filters were introduced by Saks and Seshadhri (SICOMP 2010) and the relaxed definition we study is due to Bhattacharyya et al.(RANDOM 2010), except that we further relax it by allowing additive error. Local filters for Lipschitz and monotone functions have applications to areas such as data privacy.We show that every local filter for Lipschitz or monotone functions runs in time exponential in the dimension d, even when the filter is allowed significant additive error. Prior lower bounds (for local filters with no additive error, i.e., with a = 0) applied only to more restrictive class of filters, e.g., nonadaptive filters. To prove our lower bounds, we construct families of hard functions and show that lookups of a local filter on these functions are captured by a combinatorial object that we call a c-connector. Then we present a lower bound on the maximum outdegree of a c-connector, and show that it implies the desired bounds on the running time of local filters. Our lower bounds, in particular, imply the same bound on the running time for a class of privacy mechanisms.

Complexity of pattern classes and the Lipschitz property

The primary problem in property testing is to decide whether a given function satisfies a certain property, or is far from any function satisfying it. This crucially requires a notion of distance between functions. The most prevalent notion is the Hamming distance over the {\em uniform} distribution on the domain. This restriction to uniformity is more a matter of convenience than of necessity, and it is important to investigate distances induced by more general distributions. In this paper, we make significant strides in this direction. We give simple and optimal testers for {\em bounded derivative properties} over {\em arbitrary product distributions}. Bounded derivative properties include fundamental properties such as monotonicity and Lipschitz continuity. Our results subsume almost all known results (upper and lower bounds) on monotonicity and Lipschitz testing. We prove an intimate connection between bounded derivative property testing and binary search trees (BSTs). We exhibit a tester whose query complexity is the sum of expected depths of optimal BSTs for each marginal. Furthermore, we show this sum-of-depths is also a lower bound. A fundamental technical contribution of this work is an {\em optimal dimension reduction theorem} for all bounded derivative properties, which relates the distance of a function from the property to the distance of restrictions of the function to random lines. Such a theorem has been elusive even for monotonicity for the past 15 years, and our theorem is an exponential improvement to the previous best known result.

Global exponential periodicity and global exponential stability of a class of recurrent neural networks with various activation functions and time-varying delays

Extensions of vector-valued functions with preservation of derivatives

A non-linear unknown input observer (UIO) is designed for systems with non-linearities having lower bonds on their slopes. The observer is constructed such that the non-linearity in the observer error system satisfies a sector property. Based on this property rather than the Lipschitz property, an algorithm not limited by Lipschitz constant is proposed. Further, all the parameters in the proposed non-linear UIO can be calculated by solving a linear matrix inequality. The UIO is also extended to an adaptive one for a class of uncertain non-linear systems in which the arguments of non-linear terms are perturbed by unknown parameters. Finally, simulations of electric driving system are used to illustrate the effectiveness of the UIOs.

Techniques known as Nonlinear Set Membership prediction, Kinky Inference or Lipschitz Interpolation are fast and numerically robust approaches to nonparametric machine learning that have been proposed to be utilised in the context of system identification and learning-based control. They utilise presupposed Lipschitz properties in order to compute inferences over unobserved function values. Unfortunately, most of these approaches rely on exact knowledge about the input space metric as well as about the Lipschitz constant. Furthermore, existing techniques to estimate the Lipschitz constants from the data are not robust to noise or seem to be ad-hoc and typically are decoupled from the ultimate learning and prediction task. To overcome these limitations, we propose an approach for optimising parameters of the presupposed metrics by minimising validation set prediction errors. To avoid poor performance due to local minima, we propose to utilise Lipschitz properties of the optimisation objective to ensure global optimisation success. The resulting approach is a new flexible method for nonparametric black-box learning. We illustrate its competitiveness on a set of benchmark problems.

In this paper, we define some new hyperbolic classes of functions, which will be called Bα,log* and Flog*(p,q,s) classes. Then, we introduce D-metrics in these defined hyperbolic classes. Also, these classes are shown to be complete metric spaces with respect to the corresponding metrics. Moreover, necessary and sufficient conditions are given for a composition operator Cφ to be bounded and compact from Bα,log* to Flog*(p,q,s).

Focusing on property testing tasks that have query complexity that is independent of the size of the tested object (i.e., depends on the proximity parameter only), we prove the existence of a rich hierarchy of the corresponding complexity classes. That is, for essentially any function $$q : (0, 1] \rightarrow \mathbb{N}$$ , we prove the existence of properties for which $$\epsilon$$ -testing has query complexity $$\Theta(q(\Theta(\epsilon)))$$ . Such results are proved in three standard domains that are often considered in property testing: generic functions, adjacency predicates describing (dense) graphs, and incidence functions describing bounded-degree graphs. These results complement hierarchy theorems of Goldreich, Krivelevich, Newman, and Rozenberg (Computational Complexity, 2012), which refer to the dependence of the query complexity on the size of the tested object, and focus on the case that the proximity parameter is set to some small positive constant. We actually combine both flavors and get tight results on the query complexity of testing when allowing the query complexity to depend on both the size of the object and the proximity parameter.

On fully operator Lipschitz functions

The goal of property testing is to quickly distinguish between objects which satisfy a property and objects that are ϵ-far from satisfying the property. There are now several general results in this area which show that natural properties of combinatorial objects can be tested with constant query complexity, depending only on ϵ and the property, and not on the size of the object being tested. The upper bound on the query complexity coming from the proof techniques are often enormous and impractical. It remains a major open problem if better bounds hold.Hoppen, Kohayakawa, Moreira, and Sampaio conjectured and Klimosova and Kral' proved that hereditary permutation properties are strongly testable, i.e., can be tested with respect to Kendall's tau distance. The query complexity bound coming from this proof is huge. Even for testing a single forbidden subpermutation it is of Ackermann-type in 1/ϵ. We give a new proof which gives a polynomial bound in 1/ϵ for testing a single forbidden subpermutation.Maybe surprisingly, for testing with respect to the rectangular distance, we prove there is a universal (not depending on the property), polynomial in 1/ϵ query complexity bound for two-sided testing hereditary properties of sufficiently large permutations. We further give a nearly linear bound with respect to a closely related metric which also depends on the smallest forbidden subpermutation for the property. Finally, we show that several different permutation metrics of interest are related to the rectangular distance, yielding similar results for testing with respect to these metrics.

We consider the problem of distribution-free testing of the class of monotone monomials and the class of monomials over n variables.0 While there are very efficient algorithms for testing a variety of functions classes when the underlying distribution is uniform, designing distribution-free algorithms (which must work under any arbitrary and unknown distribution), tends to be a more challenging task. When the underlying distribution is uniform, Parnas et al. (SIAM Journal on Discrete Math, 2002) give an algorithm for testing (monotone) monomials whose query complexity does not depends on n, and whose dependence on the distance parameter is (inverse) linear. In contrast, Glasner and Servedio (in Proceedings of RANDOM, 2007) prove that every distribution-free testing algorithm for monotone monomials as well as for general monomials must have query complexity Ω(n1/5) (for a constant distance parameter e). In this paper we present distribution-free testing algorithms for these classes where the query complexity of the algorithms is O(n1/2/e). We note that as opposed to previous results for distribution-free testing, our algorithms do not build on the algorithms that work under the uniform distribution. Rather, we define and exploit certain structural properties of monomials (and functions that differ from them in a nonnegligible manner), which were not used in previous work on property testing.

In the paper we investigate the regularity of the value function representing Hamilton- Jacobi equation: �Ut + H(t,x,U,�Ux) = 0 with a final condition: U(T,x )= g(x). Hamilton-Jacobi equation, in which the Hamiltonian H depends on the value of solution U, is represented by the value function with more complicated structure than the value function in Bolza problem. This function is described with the use of some class of Mayer problems related to the optimal control theory and the calculus of variation. In the paper we prove that absolutely continuous functions that are solutions of Mayer problem satisfy the Lipschitz condition. Using this fact we show that the value function is a bilateral solution of Hamilton-Jacobi equation. Moreover, we prove that continuity or the local Lipschitz condition of the function of final cost g is inherited by the value function. Our results allow to state the theorem about existence and uniqueness of bilateral solutions in the class of functions that are bounded from below and satisfy the local Lipschitz condition. In proving the main results we use recently derived necessary optimality conditions of Loewen-Rockafellar (P.D. Loewen and R.T. Rockafellar, SIAM J. Control Optim. 32 (1994) 442-470; P.D. Loewen and R.T. Rockafellar, SIAM J. Control Optim. 35 (1997) 2050-2069).

In this paper, we present a set of Newton‐type inequalities for n‐times differentiable convex functions using the Caputo fractional operator, extending classical results into the fractional calculus domain. Our exploration also includes the derivation of Newton‐type inequalities for various classes of functions by employing the Caputo fractional operator, thereby broadening the scope of these inequalities beyond convexity. In addition, we establish several fractional Newton‐type inequalities by using bounded functions in conjunction with fractional integrals. Furthermore, we develop specific fractional Newton‐type inequalities tailored to Lipschitzian functions. Moreover, the paper emphasizes the significance of fractional calculus in refining classical inequalities and demonstrates how the Caputo fractional operator provides a more generalized framework for addressing problems involving non‐integer order differentiation. The inclusion of bounded and Lipschitzian functions introduces additional layers of complexity, allowing for a more comprehensive analysis of function behaviors under fractional operations.

We study the behaviour of functions of self‐adjoint operators under relatively bounded and relatively trace class perturbation. We introduce and study the class of relatively operator Lipschitz functions. An essential role is played by double operator integrals. We also consider the class of resolvent Lipschitz functions. Then we obtain a trace formula in the case of relatively trace class perturbations and show that the maximal class of function for which the trace formula holds in the case of relatively trace class perturbations coincides with the class of relatively operator Lipschitz functions. Our methods also give us a new approach to the inequality for the spectral shift function in the case of relatively trace class perturbations.