arXiv subjects

Laurent Viennot

Publications and source records attributed to Laurent Viennot.

34 records · Page 2Linked to original sources

arXiv2019

Diameter computation on $H$-minor free graphs and graphs of bounded (distance) VC-dimension

We propose to study unweighted graphs of constant distance VC-dimension as a broad generalization of many graph classes for which we can compute the diameter in truly subquadratic-time. In particular for any fixed $H$, the class of $H$-minor free graphs has distance VC-dimension at most $|V(H)|-1$. Our first main result is that on graphs of distance VC-dimension at most $d$, for any fixed $k$ we can either compute the diameter or conclude that it is larger than $k$ in time $\tilde{\cal O}(k\cdot mn^{1-\varepsilon_d})$, where $\varepsilon_d \in (0;1)$ only depends on $d$. Then as a byproduct of our approach, we get the first truly subquadratic-time algorithm for constant diameter computation on all the nowhere dense graph classes. Finally, we show how to remove the dependency on $k$ for any graph class that excludes a fixed graph $H$ as a minor. More generally, our techniques apply to any graph with constant distance VC-dimension and polynomial expansion. As a result for all such graphs one obtains a truly subquadratic-time algorithm for computing their diameter. Our approach is based on the work of Chazelle and Welzl who proved the existence of spanning paths with strongly sublinear stabbing number for every hypergraph of constant VC-dimension. We show how to compute such paths efficiently by combining the best known approximation algorithms for the stabbing number problem with a clever use of $\varepsilon$-nets, region decomposition and other partition techniques.

cs.DS↗

arXiv2019

A Comparative Study of Neural Network Compression

There has recently been an increasing desire to evaluate neural networks locally on computationally-limited devices in order to exploit their recent effectiveness for several applications; such effectiveness has nevertheless come together with a considerable increase in the size of modern neural networks, which constitute a major downside in several of the aforementioned computationally-limited settings. There has thus been a demand of compression techniques for neural networks. Several proposal in this direction have been made, which famously include hashing-based methods and pruning-based ones. However, the evaluation of the efficacy of these techniques has so far been heterogeneous, with no clear evidence in favor of any of them over the others. The goal of this work is to address this latter issue by providing a comparative study. While most previous studies test the capability of a technique in reducing the number of parameters of state-of-the-art networks , we follow [CWT + 15] in evaluating their performance on basic ar-chitectures on the MNIST dataset and variants of it, which allows for a clearer analysis of some aspects of their behavior. To the best of our knowledge, we are the first to directly compare famous approaches such as HashedNet, Optimal Brain Damage (OBD), and magnitude-based pruning with L1 and L2 regularization among them and against equivalent-size feed-forward neural networks with simple (fully-connected) and structural (convolutional) neural networks. Rather surprisingly, our experiments show that (iterative) pruning-based methods are substantially better than the HashedNet architecture, whose compression doesn't appear advantageous to a carefully chosen convolutional network. We also show that, when the compression level is high, the famous OBD pruning heuristics deteriorates to the point of being less efficient than simple magnitude-based techniques.

cs.LG↗

arXiv2019

Hardness of Exact Distance Queries in Sparse Graphs Through Hub Labeling

A distance labeling scheme is an assignment of bit-labels to the vertices of an undirected, unweighted graph such that the distance between any pair of vertices can be decoded solely from their labels. An important class of distance labeling schemes is that of hub labelings, where a node $v \in G$ stores its distance to the so-called hubs $S_v \subseteq V$, chosen so that for any $u,v \in V$ there is $w \in S_u \cap S_v$ belonging to some shortest $uv$ path. Notice that for most existing graph classes, the best distance labelling constructions existing use at some point a hub labeling scheme at least as a key building block. Our interest lies in hub labelings of sparse graphs, i.e., those with $|E(G)| = O(n)$, for which we show a lowerbound of $\frac{n}{2^{O(\sqrt{\log n})}}$ for the average size of the hubsets. Additionally, we show a hub-labeling construction for sparse graphs of average size $O(\frac{n}{RS(n)^{c}})$ for some $0 < c < 1$, where $RS(n)$ is the so-called Ruzsa-Szemer{é}di function, linked to structure of induced matchings in dense graphs. This implies that further improving the lower bound on hub labeling size to $\frac{n}{2^{(\log n)^{o(1)}}}$ would require a breakthrough in the study of lower bounds on $RS(n)$, which have resisted substantial improvement in the last 70 years. For general distance labeling of sparse graphs, we show a lowerbound of $\frac{1}{2^{O(\sqrt{\log n})}} SumIndex(n)$, where $SumIndex(n)$ is the communication complexity of the Sum-Index problem over $Z_n$. Our results suggest that the best achievable hub-label size and distance-label size in sparse graphs may be $Θ(\frac{n}{2^{(\log n)^c}})$ for some $0<c < 1$.

cs.DS↗

arXiv2019

Fast Public Transit Routing with Unrestricted Walking through Hub Labeling

We propose a novel technique for answering routing queries in public transportation networks that allows unrestricted walking. We consider several types of queries: earliest arrival time, Pareto-optimal journeys regarding arrival time, number of transfers and walking time, and profile, i.e. finding all Pareto-optimal journeys regarding travel time and arrival time in a given time interval. Our techniques uses hub labeling to represent unlimited foot transfers and can be adapted to both classical algorithms RAPTOR and CSA. We obtain significant speedup compared to the state-of-the-art approach based on contraction hierarchies.

cs.NI↗

arXiv2019

Exploiting Hopsets: Improved Distance Oracles for Graphs of Constant Highway Dimension and Beyond

For fixed $h \geq 2$, we consider the task of adding to a graph $G$ a set of weighted shortcut edges on the same vertex set, such that the length of a shortest $h$-hop path between any pair of vertices in the augmented graph is exactly the same as the original distance between these vertices in $G$. A set of shortcut edges with this property is called an exact $h$-hopset and may be applied in processing distance queries on graph $G$. In particular, a $2$-hopset directly corresponds to a distributed distance oracle known as a hub labeling. In this work, we explore centralized distance oracles based on $3$-hopsets and display their advantages in several practical scenarios. In particular, for graphs of constant highway dimension, and more generally for graphs of constant skeleton dimension, we show that $3$-hopsets require exponentially fewer shortcuts per node than any previously described distance oracle while incurring only a quadratic increase in the query decoding time, and actually offer a speedup when compared to simple oracles based on a direct application of $2$-hopsets. Finally, we consider the problem of computing minimum-size $h$-hopset (for any $h \geq 2$) for a given graph $G$, showing a polylogarithmic-factor approximation for the case of unique shortest path graphs. When $h=3$, for a given bound on the space used by the distance oracle, we provide a construction of hopsets achieving polylog approximation both for space and query time compared to the optimal $3$-hopset oracle given the space bound.

cs.DS↗

arXiv2019

Independent Lazy Better-Response Dynamics on Network Games

We study an independent best-response dynamics on network games in which the nodes (players) decide to revise their strategies independently with some probability. We provide several bounds on the convergence time to an equilibrium as a function of this probability, the degree of the network, and the potential of the underlying games. These dynamics are somewhat more suitable for distributed environments than the classical better- and best-response dynamics where players revise their strategies "sequentially", i.e., no two players revise their strategies simultaneously.

cs.GT↗

arXiv2018

Efficient Loop Detection in Forwarding Networks and Representing Atoms in a Field of Sets

The problem of detecting loops in a forwarding network is known to be NP-complete when general rules such as wildcard expressions are used. Yet, network analyzer tools such as Netplumber (Kazemian et al., NSDI'13) or Veriflow (Khurshid et al., NSDI'13) efficiently solve this problem in networks with thousands of forwarding rules. In this paper, we complement such experimental validation of practical heuristics with the first provably efficient algorithm in the context of general rules. Our main tool is a canonical representation of the atoms (i.e. the minimal non-empty sets) of the field of sets generated by a collection of sets. This tool is particularly suited when the intersection of two sets can be efficiently computed and represented. In the case of forwarding networks, each forwarding rule is associated with the set of packet headers it matches. The atoms then correspond to classes of headers with same behavior in the network. We propose an algorithm for atom computation and provide the first polynomial time algorithm for loop detection in terms of number of classes (which can be exponential in general). This contrasts with previous methods that can be exponential, even in simple cases with linear number of classes. Second, we introduce a notion of network dimension captured by the overlapping degree of forwarding rules. The values of this measure appear to be very low in practice and constant overlapping degree ensures polynomial number of header classes. Forwarding loop detection is thus polynomial in forwarding networks with constant overlapping degree.

cs.NI↗

arXiv2016

Beyond Highway Dimension: Small Distance Labels Using Tree Skeletons

The goal of a hub-based distance labeling scheme for a network G = (V, E) is to assign a small subset S(u) $\subseteq$ V to each node u $\in$ V, in such a way that for any pair of nodes u, v, the intersection of hub sets S(u) $\cap$ S(v) contains a node on the shortest uv-path. The existence of small hub sets, and consequently efficient shortest path processing algorithms, for road networks is an empirical observation. A theoretical explanation for this phenomenon was proposed by Abraham et al. (SODA 2010) through a network parameter they called highway dimension, which captures the size of a hitting set for a collection of shortest paths of length at least r intersecting a given ball of radius 2r. In this work, we revisit this explanation, introducing a more tractable (and directly comparable) parameter based solely on the structure of shortest-path spanning trees, which we call skeleton dimension. We show that skeleton dimension admits an intuitive definition for both directed and undirected graphs, provides a way of computing labels more efficiently than by using highway dimension, and leads to comparable or stronger theoretical bounds on hub set size.

cs.DS↗

arXiv2016

Forwarding Tables Verification through Representative Header Sets

Forwarding table verification consists in checking the distributed data-structure resulting from the forwarding tables of a network. A classical concern is the detection of loops. We study this problem in the context of software-defined networking (SDN) where forwarding rules can be arbitrary bitmasks (generalizing prefix matching) and where tables are updated by a centralized controller. Basic verification problems such as loop detection are NP-hard and most previous work solves them with heuristics or SAT solvers. We follow a different approach based on computing a representation of the header classes, i.e. the sets of headers that match the same rules. This representation consists in a collection of representative header sets, at least one for each class, and can be computed centrally in time which is polynomial in the number of classes. Classical verification tasks can then be trivially solved by checking each representative header set. In general, the number of header classes can increase exponentially with header length, but it remains polynomial in the number of rules in the practical case where rules are constituted with predefined fields where exact, prefix matching or range matching is applied in each field (e.g., IP/MAC addresses, TCP/UDP ports). We propose general techniques that work in polynomial time as long as the number of classes of headers is polynomial and that do not make specific assumptions about the structure of the sets associated to rules. The efficiency of our method rely on the fact that the data-structure representing rules allows efficient computation of intersection, cardinal and inclusion. Finally, we propose an algorithm to maintain such representation in presence of updates (i.e., rule insert/update/removal). We also provide a local distributed algorithm for checking the absence of black-holes and a proof labeling scheme for locally checking the absence of loops.

cs.NI↗

arXiv2016

LiveRank: How to Refresh Old Datasets

This paper considers the problem of refreshing a dataset. More precisely , given a collection of nodes gathered at some time (Web pages, users from an online social network) along with some structure (hyperlinks, social relationships), we want to identify a significant fraction of the nodes that still exist at present time. The liveness of an old node can be tested through an online query at present time. We call LiveRank a ranking of the old pages so that active nodes are more likely to appear first. The quality of a LiveRank is measured by the number of queries necessary to identify a given fraction of the active nodes when using the LiveRank order. We study different scenarios from a static setting where the Liv-eRank is computed before any query is made, to dynamic settings where the LiveRank can be updated as queries are processed. Our results show that building on the PageRank can lead to efficient LiveRanks, for Web graphs as well as for online social networks.

cs.SI↗

arXiv2015

Toward more localized local algorithms: removing assumptions concerning global knowledge

Numerous sophisticated local algorithm were suggested in the literature for various fundamental problems. Notable examples are the MIS and $(Δ+1)$-coloring algorithms by Barenboim and Elkin [6], by Kuhn [22], and by Panconesi and Srinivasan [34], as well as the $O(Δ2)$-coloring algorithm by Linial [28]. Unfortunately, most known local algorithms (including, in particular, the aforementioned algorithms) are non-uniform, that is, local algorithms generally use good estimations of one or more global parameters of the network, e.g., the maximum degree $Δ$ or the number of nodes n. This paper provides a method for transforming a non-uniform local algorithm into a uniform one. Furthermore , the resulting algorithm enjoys the same asymp-totic running time as the original non-uniform algorithm. Our method applies to a wide family of both deterministic and randomized algorithms. Specifically, it applies to almost all state of the art non-uniform algorithms for MIS and Maximal Matching, as well as to many results concerning the coloring problem. (In particular, it applies to all aforementioned algorithms.) To obtain our transformations we introduce a new distributed tool called pruning algorithms, which we believe may be of independent interest.

cs.DC↗

arXiv2015

Self-Organizing Flows in Social Networks

Social networks offer users new means of accessing information, essentially relying on "social filtering", i.e. propagation and filtering of information by social contacts. The sheer amount of data flowing in these networks, combined with the limited budget of attention of each user, makes it difficult to ensure that social filtering brings relevant content to the interested users. Our motivation in this paper is to measure to what extent self-organization of the social network results in efficient social filtering. To this end we introduce flow games, a simple abstraction that models network formation under selfish user dynamics, featuring user-specific interests and budget of attention. In the context of homogeneous user interests, we show that selfish dynamics converge to a stable network structure (namely a pure Nash equilibrium) with close-to-optimal information dissemination. We show in contrast, for the more realistic case of heterogeneous interests, that convergence, if it occurs, may lead to information dissemination that can be arbitrarily inefficient, as captured by an unbounded "price of anarchy". Nevertheless the situation differs when users' interests exhibit a particular structure, captured by a metric space with low doubling dimension. In that case, natural autonomous dynamics converge to a stable configuration. Moreover, users obtain all the information of interest to them in the corresponding dissemination, provided their budget of attention is logarithmic in the size of their interest set.

cs.SI↗

arXiv2011

Node-Disjoint Multipath Spanners and their Relationship with Fault-Tolerant Spanners

Motivated by multipath routing, we introduce a multi-connected variant of spanners. For that purpose we introduce the $p$-multipath cost between two nodes $u$ and $v$ as the minimum weight of a collection of $p$ internally vertex-disjoint paths between $u$ and $v$. Given a weighted graph $G$, a subgraph $H$ is a $p$-multipath $s$-spanner if for all $u,v$, the $p$-multipath cost between $u$ and $v$ in $H$ is at most $s$ times the $p$-multipath cost in $G$. The $s$ factor is called the stretch. Building upon recent results on fault-tolerant spanners, we show how to build $p$-multipath spanners of constant stretch and of $\tO(n^{1+1/k})$ edges, for fixed parameters $p$ and $k$, $n$ being the number of nodes of the graph. Such spanners can be constructed by a distributed algorithm running in $O(k)$ rounds. Additionally, we give an improved construction for the case $p=k=2$. Our spanner $H$ has $O(n^{3/2})$ edges and the $p$-multipath cost in $H$ between any two node is at most twice the corresponding one in $G$ plus $O(W)$, $W$ being the maximum edge weight.

cs.NI↗

arXiv2008

Scalable Distributed Video-on-Demand: Theoretical Bounds and Practical Algorithms

We analyze a distributed system where n nodes called boxes store a large set of videos and collaborate to serve simultaneously n videos or less. We explore under which conditions such a system can be scalable while serving any sequence of demands. We model this problem through a combination of two algorithms: a video allocation algorithm and a connection scheduling algorithm. The latter plays against an adversary that incrementally proposes video requests.

cs.NI↗

arXiv2007

Acyclic Preference Systems in P2P Networks

In this work we study preference systems natural for the Peer-to-Peer paradigm. Most of them fall in three categories: global, symmetric and complementary. All these systems share an acyclicity property. As a consequence, they admit a stable (or Pareto efficient) configuration, where no participant can collaborate with better partners than their current ones. We analyze the representation of the such preference systems and show that any acyclic system can be represented with a symmetric mark matrix. This gives a method to merge acyclic preference systems and retain the acyclicity. We also consider such properties of the corresponding collaboration graph, as clustering coefficient and diameter. In particular, studying the example of preferences based on real latency measurements, we observe that its stable configuration is a small-world graph.

cs.DS↗

arXiv2006

On Using Matching Theory to Understand P2P Network Design

This paper aims to provide insight into stability of collaboration choices in P2P networks. We study networks where exchanges between nodes are driven by the desire to receive the best service available. This is the case for most existing P2P networks. We explore an evolution model derived from stable roommates theory that accounts for heterogeneity between nodes. We show that most P2P applications can be modeled using stable matching theory. This is the case whenever preference lists can be deduced from the exchange policy. In many cases, the preferences lists are characterized by an interesting acyclic property. We show that P2P networks with acyclic preferences possess a unique stable state with good convergence properties.

cs.NI↗