arXiv subjects

Nathanaël Fijalkow

Publications and source records attributed to Nathanaël Fijalkow.

At least 37 records · Page 2Linked to original sources

arXiv2022

The Theory of Universal Graphs for Infinite Duration Games

We introduce the notion of universal graphs as a tool for constructing algorithms solving games of infinite duration such as parity games and mean payoff games. In the first part we develop the theory of universal graphs, with two goals: showing an equivalence and normalisation result between different recently introduced related models, and constructing generic value iteration algorithms for any positionally determined objective. In the second part we give four applications: to parity games, to mean payoff games, to a disjunction between a parity and a mean payoff objective, and to disjunctions of several mean payoff objectives. For each of these four cases we construct algorithms achieving or improving over the best known time and space complexity.

cs.FL↗

arXiv2022

Probabilistic Automata of Bounded Ambiguity

Probabilistic automata are an extension of nondeterministic finite automata in which transitions are annotated with probabilities. Despite its simplicity, this model is very expressive and many of the associated algorithmic questions are undecidable. In this work we focus on the emptiness problem (and its variant the value problem), which asks whether a given probabilistic automaton accepts some word with probability greater than a given threshold. We consider a natural and well-studied structural restriction on automata, namely the degree of ambiguity, which is defined as the maximum number of accepting runs over all words. The known undecidability proofs exploits infinite ambiguity and so we focus on the case of finitely ambiguous probabilistic automata. Our main contributions are to construct efficient algorithms for analysing finitely ambiguous probabilistic automata through a reduction to a multi-objective optimisation problem called the stochastic path problem. We obtain a polynomial time algorithm for approximating the value of probabilistic automata of fixed ambiguity and a quasi-polynomial time algorithm for the emptiness problem for 2-ambiguous probabilistic automata. We complement these positive results by an inapproximability result stating that the value of finitely ambiguous probabilistic automata cannot be approximated unless PTIME = NP.

cs.FL↗

arXiv2021

Controlling a random population

Bertrand et al. introduced a model of parameterised systems, where each agent is represented by a finite state system, and studied the following control problem: for any number of agents, does there exist a controller able to bring all agents to a target state? They showed that the problem is decidable and EXPTIME-complete in the adversarial setting, and posed as an open problem the stochastic setting, where the agent is represented by a Markov decision process. In this paper, we show that the stochastic control problem is decidable. Our solution makes significant uses of well quasi orders, of the max-flow min-cut theorem, and of the theory of regular cost functions. We introduce an intermediate problem of independence interest called the sequential flow problem and study its complexity.

cs.FL↗

arXiv2021

Scaling Neural Program Synthesis with Distribution-based Search

We consider the problem of automatically constructing computer programs from input-output examples. We investigate how to augment probabilistic and neural program synthesis methods with new search algorithms, proposing a framework called distribution-based search. Within this framework, we introduce two new search algorithms: Heap Search, an enumerative method, and SQRT Sampling, a probabilistic method. We prove certain optimality guarantees for both methods, show how they integrate with probabilistic and neural techniques, and demonstrate how they can operate at scale across parallel compute environments. Collectively these findings offer theoretical and applied studies of search algorithms for program synthesis that integrate with recent developments in machine-learned program synthesizers.

cs.LG↗

arXiv2021

New Algorithms for Combinations of Objectives using Separating Automata

The notion of separating automata was introduced by Bojanczyk and Czerwinski for understanding the first quasipolynomial time algorithm for parity games. In this paper we show that separating automata is a powerful tool for constructing algorithms solving games with combinations of objectives. We construct two new algorithms: the first for disjunctions of parity and mean payoff objectives, matching the best known complexity, and the second for disjunctions of mean payoff objectives, improving on the state of the art. In both cases the algorithms are obtained through the construction of small separating automata, using as black boxes the existing constructions for parity objectives and for mean payoff objectives.

cs.GT↗

arXiv2021

The Complexity of Learning Linear Temporal Formulas from Examples

In this paper we initiate the study of the computational complexity of learning linear temporal logic (LTL) formulas from examples. We construct approximation algorithms for fragments of LTL and prove hardness results; in particular we obtain tight bounds for approximation of the fragment containing only the next operator and conjunctions, and prove NP-completeness results for many fragments.

cs.FL↗

arXiv2020

Alternating Tree Automata with Qualitative Semantics

We study alternating automata with qualitative semantics over infinite binary trees: alternation means that two opposing players construct a decoration of the input tree called a run, and the qualitative semantics says that a run of the automaton is accepting if almost all branches of the run are accepting. In this paper we prove a positive and a negative result for the emptiness problem of alternating automata with qualitative semantics. The positive result is the decidability of the emptiness problem for the case of Büchi acceptance condition. An interesting aspect of our approach is that we do not extend the classical solution for solving the emptiness problem of alternating automata, which first constructs an equivalent non-deterministic automaton. Instead, we directly construct an emptiness game making use of imperfect information. The negative result is the undecidability of the emptiness problem for the case of co-Büchi acceptance condition. This result has two direct consequences: the undecidability of monadic second-order logic extended with the qualitative path-measure quantifier, and the undecidability of the emptiness problem for alternating tree automata with non-zero semantics, a recently introduced probabilistic model of alternating tree automata.

cs.FL↗

arXiv2020

Trace Refinement in Labelled Markov Decision Processes

Given two labelled Markov decision processes (MDPs), the trace-refinement problem asks whether for all strategies of the first MDP there exists a strategy of the second MDP such that the induced labelled Markov chains are trace-equivalent. We show that this problem is decidable in polynomial time if the second MDP is a Markov chain. The algorithm is based on new results on a particular notion of bisimulation between distributions over the states. However, we show that the general trace-refinement problem is undecidable, even if the first MDP is a Markov chain. Decidability of those problems was stated as open in 2008. We further study the decidability and complexity of the trace-refinement problem provided that the strategies are restricted to be memoryless.

cs.LO↗

arXiv2019

Lower bounds for the state complexity of probabilistic languages and the language of prime numbers

This paper studies the complexity of languages of finite words using automata theory. To go beyond the class of regular languages, we consider infinite automata and the notion of state complexity defined by Karp. Motivated by the seminal paper of Rabin from 1963 introducing probabilistic automata, we study the (deterministic) state complexity of probabilistic languages and prove that probabilistic languages can have arbitrarily high deterministic state complexity. We then look at alternating automata as introduced by Chandra, Kozen and Stockmeyer: such machines run independent computations on the word and gather their answers through boolean combinations. We devise a lower bound technique relying on boundedly generated lattices of languages, and give two applications of this technique. The first is a hierarchy theorem, stating that there are languages of arbitrarily high polynomial alternating state complexity, and the second is a linear lower bound on the alternating state complexity of the prime numbers written in binary. This second result strengthens a result of Hartmanis and Shank from 1968, which implies an exponentially worse lower bound for the same model.

cs.FL↗

arXiv2019

Data Generation for Neural Programming by Example

Programming by example is the problem of synthesizing a program from a small set of input / output pairs. Recent works applying machine learning methods to this task show promise, but are typically reliant on generating synthetic examples for training. A particular challenge lies in generating meaningful sets of inputs and outputs, which well-characterize a given program and accurately demonstrate its behavior. Where examples used for testing are generated by the same method as training data then the performance of a model may be partly reliant on this similarity. In this paper we introduce a novel approach using an SMT solver to synthesize inputs which cover a diverse set of behaviors for a given program. We carry out a case study comparing this method to existing synthetic data generation procedures in the literature, and find that data generated using our approach improves both the discriminatory power of example sets and the ability of trained machine learning models to generalize to unfamiliar data.

cs.LG↗

arXiv2019

Verification of Neural Networks: Specifying Global Robustness using Generative Models

The success of neural networks across most machine learning tasks and the persistence of adversarial examples have made the verification of such models an important quest. Several techniques have been successfully developed to verify robustness, and are now able to evaluate neural networks with thousands of nodes. The main weakness of this approach is in the specification: robustness is asserted on a validation set consisting of a finite set of examples, i.e. locally. We propose a notion of global robustness based on generative models, which asserts the robustness on a very large and representative set of examples. We show how this can be used for verifying neural networks. In this paper we experimentally explore the merits of this approach, and show how it can be used to construct realistic adversarial examples.

cs.LG↗

arXiv2019

A Robust Class of Linear Recurrence Sequences

We introduce a subclass of linear recurrence sequences which we call poly-rational sequences because they are denoted by rational expressions closed under sum and product. We show that this class is robust by giving several characterisations: polynomially ambiguous weighted automata, copyless cost-register automata, rational formal series, and linear recurrence sequences whose eigenvalues are roots of rational numbers.

cs.FL↗

arXiv2019

On the Monniaux Problem in Abstract Interpretation

The Monniaux Problem in abstract interpretation asks, roughly speaking, whether the following question is decidable: given a program $P$, a safety (\emph{e.g.}, non-reachability) specification $φ$, and an abstract domain of invariants $\mathcal{D}$, does there exist an inductive invariant $I$ in $\mathcal{D}$ guaranteeing that program $P$ meets its specification $φ$. The Monniaux Problem is of course parameterised by the classes of programs and invariant domains that one considers. In this paper, we show that the Monniaux Problem is undecidable for unguarded affine programs and semilinear invariants (unions of polyhedra). Moreover, we show that decidability is recovered in the important special case of simple linear loops.

cs.LO↗

arXiv2019

On the Decidability of Reachability in Linear Time-Invariant Systems

We consider the decidability of state-to-state reachability in linear time-invariant control systems over discrete time. We analyse this problem with respect to the allowable control sets, which in general are assumed to be defined by boolean combinations of linear inequalities. Decidability of the version of the reachability problem in which control sets are affine subspaces of $\mathbb{R}^n$ is a fundamental result in control theory. Our first result is that reachability is undecidable if the set of controls is a finite union of affine subspaces. We also consider versions of the reachability problem in which (i)~the set of controls consists of a single affine subspace together with the origin and (ii)~the set of controls is a convex polytope. In these two cases we respectively show that the reachability problem is as hard as Skolem's Problem and the Positivity Problem for linear recurrence sequences (whose decidability has been open for several decades). Our main contribution is to show decidability of a version of the reachability problem in which control sets are convex polytopes, under certain spectral assumptions on the transition matrix.

math.OC↗

arXiv2019

The complexity of mean payoff games using universal graphs

We study the computational complexity of solving mean payoff games. This class of games can be seen as an extension of parity games, and they have similar complexity status: in both cases solving them is in $\textbf{NP} \cap \textbf{coNP}$ and not known to be in $\textbf{P}$. In a breakthrough result Calude, Jain, Khoussainov, Li, and Stephan constructed in 2017 a quasipolynomial time algorithm for solving parity games, which was quickly followed by two other algorithms with the same complexity. Our objective is to investigate how these techniques can be extended to the study of mean payoff games. The starting point is the notion of separating automata, which has been used to present all three quasipolynomial time algorithms for parity games and gives the best complexity to date. The notion naturally extends to mean payoff games and yields a class of algorithms for solving mean payoff games. The contribution of this paper is to prove tight bounds on the complexity of algorithms in this class. We construct two new algorithms for solving mean payoff games. Our first algorithm depends on the largest weight $N$ (in absolute value) appearing in the graph and runs in sublinear time in $N$, improving over the previously known linear dependence in $N$. Our second algorithm runs in polynomial time for a fixed number $k$ of weights. We complement our upper bounds by providing in both cases almost matching lower bounds, showing the limitations of the separating automata approach. We show that we cannot hope to improve on the dependence in $N$ nor break the linear dependence in the exponent in the number $k$ of weights. In particular, this shows that separating automata do not yield a quasipolynomial algorithm for solving mean payoff games.

cs.GT↗

arXiv2018

Universal trees grow inside separating automata: Quasi-polynomial lower bounds for parity games

Several distinct techniques have been proposed to design quasi-polynomial algorithms for solving parity games since the breakthrough result of Calude, Jain, Khoussainov, Li, and Stephan (2017): play summaries, progress measures and register games. We argue that all those techniques can be viewed as instances of the separation approach to solving parity games, a key technical component of which is constructing (explicitly or implicitly) an automaton that separates languages of words encoding plays that are (decisively) won by either of the two players. Our main technical result is a quasi-polynomial lower bound on the size of such separating automata that nearly matches the current best upper bounds. This forms a barrier that all existing approaches must overcome in the ongoing quest for a polynomial-time algorithm for solving parity games. The key and fundamental concept that we introduce and study is a universal ordered tree. The technical highlights are a quasi-polynomial lower bound on the size of universal ordered trees and a proof that every separating safety automaton has a universal tree hidden in its state space.

cs.FL↗

arXiv2018

Parity games and universal graphs

This paper is a contribution to the study of parity games and the recent constructions of three quasipolynomial time algorithms for solving them. We revisit a result of Czerwiński, Daviaud, Fijalkow, Jurdziński, Lazić, and Parys witnessing a quasipolynomial barrier for all three quasipolynomial time algorithms. The argument is that all three algorithms can be understood as constructing a so-called separating automaton, and to give a quasipolynomial lower bond on the size of separating automata. We give an alternative proof of this result. The key innovations of this paper are the notion of universal graphs and the idea of saturation.

cs.GT↗

arXiv2018

An Optimal Value Iteration Algorithm for Parity Games

The quest for a polynomial time algorithm for solving parity games gained momentum in 2017 when two different quasipolynomial time algorithms were constructed. In this paper, we further analyse the second algorithm due to Jurdziński and Lazić and called the succinct progress measure algorithm. It was presented as an improvement over a previous algorithm called the small progress measure algorithm, using a better data structure. The starting point of this paper is the observation that the underlying data structure for both progress measure algorithms are (subgraph-)universal trees. We show that in fact any universal tree gives rise to a value iteration algorithm à la succinct progress measure, and the complexity of the algorithm is proportional to the size of the chosen universal tree. We then show that both algorithms are instances of this generic algorithm for two constructions of universal trees, the first of exponential size (for small progress measure) and the second of quasipolynomial size (for succinct progress measure). The technical result of this paper is to show that the latter construction is asymptotically tight: universal trees have at least quasipolynomial size. This suggests that the succinct progress measure algorithm of Jurdziński and Lazić is in this framework optimal, and that the polynomial time algorithm for parity games is hiding someplace else.

cs.GT↗