arXiv subjects

Stephan Mertens

Publications and source records attributed to Stephan Mertens.

At least 37 records · Page 2Linked to original sources

arXiv2011

Parallel Complexity of Random Boolean Circuits

Random instances of feedforward Boolean circuits are studied both analytically and numerically. Evaluating these circuits is known to be a P-complete problem and thus, in the worst case, believed to be impossible to perform, even given a massively parallel computer, in time much less than the depth of the circuit. Nonetheless, it is found that for some ensembles of random circuits, saturation to a fixed truth value occurs rapidly so that evaluation of the circuit can be accomplished in much less parallel time than the depth of the circuit. For other ensembles saturation does not occur and circuit evaluation is apparently hard. In particular, for some random circuits composed of connectives with five or more inputs, the number of true outputs at each level is a chaotic sequence. Finally, while the average case complexity depends on the choice of ensemble, it is shown that for all ensembles it is possible to simultaneously construct a typical circuit together with its solution in polylogarithmic parallel time.

cond-mat.dis-nn↗

arXiv2009

Asymptotics of Lagged Fibonacci Sequences

Consider "lagged" Fibonacci sequences $a(n) = a(n-1)+a(\lfloor n/k\rfloor)$ for $k > 1$. We show that $\lim_{n\to\infty} a(kn)/a(n)\cdot\ln n/n = k\ln k$ and we demonstrate the slow numerical convergence to this limit and how to deal with this slow convergence. We also discuss the connection between two classical results of N.G. de Bruijn and K. Mahler on the asymptotics of $a(n)$.

math.CO↗

arXiv2009

Random Number Generators: A Survival Guide for Large Scale Simulations

Monte Carlo simulations are an important tool in statistical physics, complex systems science, and many other fields. An increasing number of these simulations is run on parallel systems ranging from multicore desktop computers to supercomputers with thousands of CPUs. This raises the issue of generating large amounts of random numbers in a parallel application. In this lecture we will learn just enough of the theory of pseudo random number generation to make wise decisions on how to choose and how to use random number generators when it comes to large scale, parallel simulations.

cond-mat.stat-mech↗

arXiv2008

Brachistochrones With Loose Ends

The classical problem of the brachistochrone asks for the curve down which a body sliding from rest and accelerated by gravity will slip (without friction) from one point to another in least time. In undergraduate courses on classical mechanics, the solution of this problem is the primary example of the power of the variational calculus. Here we address the generalized brachistochrone problem that asks for the fastest sliding curve between a point and a given curve or between two given curves. The generalized problem can be solved by considering variations with varying endpoints. We will contrast the formal solution with a much simpler solution based on symmetry and kinematic reasoning. Our exposition should encourage teachers to include variational problems with free boundary conditions in their courses and students to try simple, intuitive solutions first.

physics.class-ph↗

arXiv2008

Analysis of the Karmarkar-Karp Differencing Algorithm

The Karmarkar-Karp differencing algorithm is the best known polynomial time heuristic for the number partitioning problem, fundamental in both theoretical computer science and statistical physics. We analyze the performance of the differencing algorithm on random instances by mapping it to a nonlinear rate equation. Our analysis reveals strong finite size effects that explain why the precise asymptotics of the differencing solution is hard to establish by simulations. The asymptotic series emerging from the rate equation satisfies all known bounds on the Karmarkar-Karp algorithm and projects a scaling $n^{-c\ln n}$, where $c=1/(2\ln2)=0.7213...$. Our calculations reveal subtle relations between the algorithm and Fibonacci-like sequences, and we establish an explicit identity to that effect.

math.NA↗

arXiv2007

Random numbers for large scale distributed Monte Carlo simulations

Monte Carlo simulations are one of the major tools in statistical physics, complex system science, and other fields, and an increasing number of these simulations is run on distributed systems like clusters or grids. This raises the issue of generating random numbers in a parallel, distributed environment. In this contribution we demonstrate that multiple linear recurrences in finite fields are an ideal method to produce high quality pseudorandom numbers in sequential and parallel algorithms. Their known weakness (failure of sampling points in high dimensions) can be overcome by an appropriate delinearization that preserves all desirable properties of the underlying linear sequence.

cond-mat.other↗

arXiv2005

Random Stable Matchings

The stable matching problem is a prototype model in economics and social sciences where agents act selfishly to optimize their own satisfaction, subject to mutually conflicting constraints. A stable matching is a pairing of adjacent vertices in a graph such that no unpaired vertices prefer each other to their partners under the matching. The problem of finding stable matchings is known as stable marriage problem (on bipartite graphs) or as stable roommates problem (on the complete graph). It is well-known that not all instances on non-bipartite graphs admit a stable matching. Here we present numerical results for the probability that a graph with $n$ vertices and random preference relations admits a stable matching. In particular we find that this probability decays algebraically on graphs with connectivity $Θ(n)$ and exponentially on regular grids. On finite connectivity Erdös-Rényi graphs the probability converges to a value larger than zero. Based on the numerical results and some heuristic reasoning we formulate five conjectures on the asymptotic properties of random stable matchings.

cond-mat.dis-nn↗

arXiv2005

Proof of the local REM conjecture for number partitioning II: growing energy scales

We continue our analysis of the number partitioning problem with $n$ weights chosen i.i.d. from some fixed probability distribution with density $ρ$. In Part I of this work, we established the so-called local REM conjecture of Bauke, Franz and Mertens. Namely, we showed that, as $n \to \infty$, the suitably rescaled energy spectrum above some {\it fixed} scale $α$ tends to a Poisson process with density one, and the partitions corresponding to these energies become asymptotically uncorrelated. In this part, we analyze the number partitioning problem for energy scales $α_n$ that grow with $n$, and show that the local REM conjecture holds as long as $n^{-1/4}α_n \to 0$, and fails if $α_n$ grows like $κn^{1/4}$ with $κ>0$. We also consider the SK-spin glass model, and show that it has an analogous threshold: the local REM conjecture holds for energies of order $o(n)$, and fails if the energies grow like $κn$ with $κ>0$.

cond-mat.dis-nn↗

arXiv2005

Proof of the local REM conjecture for number partitioning I: Constant energy scales

The number partitioning problem is a classic problem of combinatorial optimization in which a set of $n$ numbers is partitioned into two subsets such that the sum of the numbers in one subset is as close as possible to the sum of the numbers in the other set. When the $n$ numbers are i.i.d. variables drawn from some distribution, the partitioning problem turns out to be equivalent to a mean-field antiferromagnetic Ising spin glass. In the spin glass representation, it is natural to define energies -- corresponding to the costs of the partitions, and overlaps -- corresponding to the correlations between partitions. Although the energy levels of this model are {\em a priori} highly correlated, a surprising recent conjecture asserts that the energy spectrum of number partitioning is locally that of a random energy model (REM): the spacings between nearby energy levels are uncorrelated. In other words, the properly scaled energies converge to a Poisson process. The conjecture also asserts that the corresponding spin configurations are uncorrelated, indicating vanishing overlaps in the spin glass representation. In this paper, we prove these two claims, collectively known as the local REM conjecture.

cond-mat.dis-nn↗

arXiv2005

Threshold values of Random K-SAT from the cavity method

Using the cavity equations of \cite{mezard:parisi:zecchina:02,mezard:zecchina:02}, we derive the various threshold values for the number of clauses per variable of the random $K$-satisfiability problem, generalizing the previous results to $K \ge 4$. We also give an analytic solution of the equations, and some closed expressions for these thresholds, in an expansion around large $K$. The stability of the solution is also computed. For any $K$, the satisfiability threshold is found to be in the stable region of the solution, which adds further credit to the conjecture that this computation gives the exact satisfiability threshold.

cs.CC↗

arXiv2004

Universality in the level statistics of disordered systems

Energy spectra of disordered systems share a common feature: if the entropy of the quenched disorder is larger than the entropy of the dynamical variables, the spectrum is locally that of a random energy model and the correlation between energy and configuration is lost. We demonstrate this effect for the Edwards-Anderson model, but we also discuss its universality.

cond-mat.dis-nn↗

arXiv2004

Entropy of Pseudo Random Number Generators

Since the work of Ferrenberg et al.[PRL 69, (1992)] some pseudo random number generators are known to yield wrong results in cluster Monte Carlo simulations. In this contribution the fundamental mechanism behind this failure is discussed. Almost all random number generators calculate a new pseudo random number $x_i$ from preceding values, $x_i = f(x_{i-1}, x_{i-2},..., x_{i-q})$. Failure of these generators in cluster Monte Carlo simulations and related experiments can be attributed to the low entropy of the production rule $f()$ conditioned on the statistics of the input values $x_{i-1},...,x_{i-q}$. Being a measure only of the arithmetic operations in the generator rule, the conditional entropy is independent of the lag in the recurrence or the period of the sequence. In that sense it measures a more profound quality of a random number generator than empirical tests with their limited horizon.

cond-mat.stat-mech↗

arXiv2004

Number partitioning as random energy model

Number partitioning is a classical problem from combinatorial optimisation. In physical terms it corresponds to a long range anti-ferromagnetic Ising spin glass. It has been rigorously proven that the low lying energies of number partitioning behave like uncorrelated random variables. We claim that neighbouring energy levels are uncorrelated almost everywhere on the energy axis, and that energetically adjacent configurations are uncorrelated, too. Apparently there is no relation between geometry (configuration) and energy that could be exploited by an optimization algorithm. This ``local random energy'' picture of number partitioning is corroborated by numerical simulations and heuristic arguments.

cond-mat.dis-nn↗

arXiv2004

Pseudo Random Coins Show More Heads Than Tails

Tossing a coin is the most elementary Monte Carlo experiment. In a computer the coin is replaced by a pseudo random number generator. It can be shown analytically and by exact enumerations that popular random number generators are not capable of imitating a fair coin: pseudo random coins show more heads than tails. This bias explains the empirically observed failure of some random number generators in random walk experiments. It can be traced down to the special role of the value zero in the algebra of finite fields.

cond-mat.stat-mech↗

arXiv2003

The Easiest Hard Problem: Number Partitioning

Number partitioning is one of the classical NP-hard problems of combinatorial optimization. It has applications in areas like public key encryption and task scheduling. The random version of number partitioning has an "easy-hard" phase transition similar to the phase transitions observed in other combinatorial problems like $k$-SAT. In contrast to most other problems, number partitioning is simple enough to obtain detailled and rigorous results on the "hard" and "easy" phase and the transition that separates them. We review the known results on random integer partitioning, give a very simple derivation of the phase transition and discuss the algorithmic implications of both phases.

cond-mat.dis-nn↗

arXiv2002

Phase Transition in Multiprocessor Scheduling

The problem of distributing the workload on a parallel computer to minimize the overall runtime is known as Multiprocessor Scheduling Problem. It is NP-hard, but like many other NP-hard problems, the average hardness of random instances displays an ``easy-hard'' phase transition. The transition in Multiprocessor Scheduling can be analyzed using elementary notions from crystallography (Bravais lattices) and statistical mechanics (Potts vectors). The analysis reveals the control parameter of the transition and its critical value including finite size corrections. The transition is identified in the performance of practical scheduling algorithms.

cond-mat.dis-nn↗

arXiv2002

Computational Complexity for Physicists

These lecture notes are an informal introduction to the theory of computational complexity and its links to quantum computing and statistical mechanics.

cond-mat.stat-mech↗

arXiv2000

A physicist's approach to number partitioning

The statistical physics approach to the number partioning problem, a classical NP-hard problem, is both simple and rewarding. Very basic notions and methods from statistical mechanics are enough to obtain analytical results for the phase boundary that separates the ``easy-to-solve'' from the ``hard-to-solve'' phase of the NPP as well as for the probability distributions of the optimal and sub-optimal solutions. In addition, it can be shown that solving a number partioning problem of size $N$ to some extent corresponds to locating the minimum in an unsorted list of $\bigo{2^N}$ numbers. Considering this correspondence it is not surprising that known heuristics for the partitioning problem are not significantly better than simple random search.

cond-mat↗