arXiv subjects

Raissa M. D'Souza

Publications and source records attributed to Raissa M. D'Souza.

49 records · Page 3Linked to original sources

arXiv2012

Efficient scheduling using complex networks

We consider the problem of efficiently scheduling the production of goods for a model steel manufacturing company. We propose a new approach for solving this classic problem, using techniques from the statistical physics of complex networks in conjunction with depth-first search to generate a successful, flexible, schedule. The schedule generated by our algorithm is more efficient and outperforms schedules selected at random from those observed in real steel manufacturing processes. Finally, we explore whether the proposed approach could be beneficial for long term planning.

physics.soc-ph↗

arXiv2011

Suppressing cascades of load in interdependent networks

Understanding how interdependence among systems affects cascading behaviors is increasingly important across many fields of science and engineering.Inspired by cascades of load shedding in coupled electric grids and other infrastructure, we study the Bak-Tang-Wiesenfeld sandpile model on modular random graphs and on graphs based on actual, interdependent power grids. Starting from two isolated networks, adding some connectivity between them is beneficial, for it suppresses the largest cascades in each system. Too much interconnectivity, however, becomes detrimental for two reasons. First, interconnections open pathways for neighboring networks to inflict large cascades. Second, as in real infrastructure, new interconnections increase capacity and total possible load, which fuels even larger cascades. Using a multitype branching process and simulations we show these effects and estimate the optimal level of interconnectivity that balances their tradeoffs. Such equilibria could allow, for example, power grid owners to minimize the largest cascades in their grid. We also show that asymmetric capacity among interdependent networks affects the optimal connectivity that each prefers and may lead to an arms race for greater capacity. Our multitype branching process framework provides building blocks for better prediction of cascading processes on modular random graphs and on multi-type networks in general.

physics.soc-ph↗

arXiv2011

Local cluster aggregation models of explosive percolation

We introduce perhaps the simplest models of graph evolution with choice that demonstrate discontinuous percolation transitions and can be analyzed via mathematical evolution equations. These models are local, in the sense that at each step of the process one edge is selected from a small set of potential edges sharing common vertices and added to the graph. We show that the evolution can be accurately described by a system of differential equations and that such models exhibit the discontinuous emergence of the giant component. Yet, they also obey scaling behaviors characteristic of continuous transitions, with scaling exponents that differ from the classic Erdos-Renyi model.

cond-mat.dis-nn↗

arXiv2011

Strongly discontinuous explosive percolation with multiple giant components

We generalize the random graph evolution process of Bohman, Frieze, and Wormald [T. Bohman, A. Frieze, and N. C. Wormald, Random Struct. Algorithms, 25, 432 (2004)]. Potential edges, sampled uniformly at random from the complete graph, are considered one at a time and either added to the graph or rejected provided that the fraction of accepted edges is never smaller than a decreasing function asymptotically approaching the value alpha = 1/2. We show that multiple giant components appear simultaneously in a strongly discontinuous percolation transition and remain distinct. Furthermore, tuning the value of alpha determines the number of such components with smaller alpha leading to an increasingly delayed and more explosive transition. The location of the critical point and strongly discontinuous nature are not affected if only edges which span components are sampled.

cond-mat.dis-nn↗

arXiv2010

Resilience and rewiring of the passenger airline networks in the United States

The air transportation network, a fundamental component of critical infrastructure, is formed from a collection of individual air carriers, each one with a methodically designed and engineered network structure. We analyze the individual structures of the seven largest passenger carriers in the USA and find that networks with dense interconnectivity, as quantified by large k-cores for high values of k, are extremely resilient to both targeted removal of airports (nodes) and random removal of flight paths paths (edges). Such networks stay connected and incur minimal increase in an heuristic travel time despite removal of a majority of nodes or edges. Similar results are obtained for targeted removal based on either node degree or centrality. We introduce network rewiring schemes that boost resilience to different levels of perturbation while preserving total number of flight and gate requirements. Recent studies have focused on the asymptotic optimality of hub-and-spoke spatial networks under normal operating conditions, yet our results indicate that point-to-point architectures can be much more resilient to perturbations.

physics.soc-ph↗

arXiv2010

Sandpile cascades on interacting tree-like networks

The vulnerability of an isolated network to cascades is fundamentally affected by its interactions with other networks. Motivated by failures cascading among electrical grids, we study the Bak-Tang-Wiesenfeld sandpile model on two sparsely-coupled random regular graphs. By approximating avalanches (cascades) as a multi-type branching process and using a generalization of Lagrange's expansion to multiple variables, we calculate the distribution of avalanche sizes within each network. Due to coupling, large avalanches in the individual networks are mitigated--in contrast to the conclusion for a simpler model [36]. Yet when compared to uncoupled networks, interdependent networks more frequently suffer avalanches that are large in both networks. Thus sparse connections between networks stabilize them individually but destabilize them jointly, as coupling introduces reservoirs for extra load yet also inflicts new stresses. These results suggest that in practice, to greedily mitigate large avalanches in one network, add connections between networks; conversely, to mitigate avalanches that are large in both networks, remove connections between networks. We also show that when only one network receives load, the largest avalanches in the second network increase in size and in frequency, an effect that is amplified with increased coupling between networks and with increased disparity in total capacity. Our framework is applicable to modular networks as well as to interacting networks and provides building blocks for better prediction of cascading processes on networks in general.

cond-mat.dis-nn↗

arXiv2009

Percolation on interacting networks

Most networks of interest do not live in isolation. Instead they form components of larger systems in which multiple networks with distinct topologies coexist and where elements distributed amongst different networks may interact directly. Here we develop a mathematical framework based on generating functions for analyzing a system of L interacting networks given the connectivity within and between networks. We derive exact expressions for the percolation threshold describing the onset of large-scale connectivity in the system of networks and each network individually. These general expressions apply to networks with arbitrary degree distributions and we explicitly evaluate them for L=2 interacting networks with a few choices of degree distributions. We show that the percolation threshold in an individual network can be significantly lowered once "hidden" connections to other networks are considered. We show applications of the framework to two real-world systems involving communications networks and socio-tecnical congruence in software systems.

cond-mat.dis-nn↗

arXiv2009

Modeling and verifying a broad array of network properties

Motivated by widely observed examples in nature, society and software, where groups of already related nodes arrive together and attach to an existing network, we consider network growth via sequential attachment of linked node groups, or graphlets. We analyze the simplest case, attachment of the three node V-graphlet, where, with probability alpha, we attach a peripheral node of the graphlet, and with probability (1-alpha), we attach the central node. Our analytical results and simulations show that tuning alpha produces a wide range in degree distribution and degree assortativity, achieving assortativity values that capture a diverse set of many real-world systems. We introduce a fifteen-dimensional attribute vector derived from seven well-known network properties, which enables comprehensive comparison between any two networks. Principal Component Analysis (PCA) of this attribute vector space shows a significantly larger coverage potential of real-world network properties by a simple extension of the above model when compared against a classic model of network growth.

cond-mat.stat-mech↗

arXiv2009

Network Growth with Feedback

Existing models of network growth typically have one or two parameters or strategies which are fixed for all times. We introduce a general framework where feedback on the current state of a network is used to dynamically alter the values of such parameters. A specific model is analyzed where limited resources are shared amongst arriving nodes, all vying to connect close to the root. We show that tunable feedback leads to growth of larger, more efficient networks. Exact results show that linear scaling of resources with system size yields crossover to a trivial condensed state, which can be considerably delayed with sublinear scaling.

physics.soc-ph↗

arXiv2008

Periodic States, Local Effects and Coexistence in the BML Traffic Jam Model

The Biham-Middleton-Levine model (BML) is simple lattice model of traffic flow, self-organization and jamming. Rather than a sharp phase transition between free-flow and jammed, it was recently shown that there is a region where stable intermediate states exist, with details dependent on the aspect ratio of the underlying lattice. Here we investigate square aspect ratios, focusing on the region where random, disordered intermediate (DI) states and conventional global jam (GJ) states coexist, and show that DI states dominate for some densities and timescales. Moreover, we show that periodic intermediate (PI) states can also coexist. PI states converge to periodic limit cycles with short recurrence times and were previously conjectured to arise from idiosyncrasies of relatively prime aspect ratios. The observed coexistence of DI, PI and GJ states shows that global parameters, density together with aspect ratio, are not sufficient to determine the full jamming outcome. We investigate additional features that lead towards jamming and show that a strategic perturbation of a few selected bits can change the nature of the flow, nucleating a global jam.

cond-mat.stat-mech↗

arXiv2007

The power of choice in network growth

The "power of choice" has been shown to radically alter the behavior of a number of randomized algorithms. Here we explore the effects of choice on models of tree and network growth. In our models each new node has k randomly chosen contacts, where k > 1 is a constant. It then attaches to whichever one of these contacts is most desirable in some sense, such as its distance from the root or its degree. Even when the new node has just two choices, i.e., when k=2, the resulting network can be very different from a random graph or tree. For instance, if the new node attaches to the contact which is closest to the root of the tree, the distribution of depths changes from Poisson to a traveling wave solution. If the new node attaches to the contact with the smallest degree, the degree distribution is closer to uniform than in a random graph, so that with high probability there are no nodes in the network with degree greater than O(log log N). Finally, if the new node attaches to the contact with the largest degree, we find that the degree distribution is a power law with exponent -1 up to degrees roughly equal to k, with an exponential cutoff beyond that; thus, in this case, we need k >> 1 to see a power law over a wide range of degrees.

cond-mat.stat-mech↗

arXiv2002

Dimension-splitting for simplifying diffusion in lattice-gas models

We introduce a simplified technique for incorporating diffusive phenomena into lattice-gas molecular dynamics models. In this method, spatial interactions take place one dimension at a time, with a separate fractional timestep devoted to each dimension, and with all dimensions treated identically. We show that the model resulting from this technique is equivalent to the macroscopic diffusion equation in the appropriate limit. This technique saves computational resources and reduces the complexity of model design, programming, debugging, simulation and analysis. For example, a reaction-diffusion simulation can be designed and tested as a one-dimensional system, and then directly extended to two or more dimensions. We illustrate the use of this approach in constructing a microscopically reversible model of diffusion limited aggregation as well as in a model of growth of biological films.

nlin.CG↗

arXiv1998

A thermodynamically reversible generalization of Diffusion Limited Aggregation

We introduce a lattice gas model of cluster growth via the diffusive aggregation of particles in a closed system obeying a local, deterministic, microscopically reversible dynamics. This model roughly corresponds to placing the irreversible Diffusion Limited Aggregation model (DLA) in contact with a heat bath. Particles release latent heat when aggregating, while singly connected cluster members can absorb heat and evaporate. The heat bath is initially empty, hence we observe the flow of entropy from the aggregating gas of particles into the heat bath, which is being populated by diffusing heat tokens. Before the population of the heat bath stabilizes, the cluster morphology (quantified by the fractal dimension) is similar to a standard DLA cluster. The cluster then gradually anneals, becoming more tenuous, until reaching configurational equilibrium when the cluster morphology resembles a quenched branched random polymer. As the microscopic dynamics is invertible, we can reverse the evolution, observe the inverse flow of heat and entropy, and recover the initial condition. This simple system provides an explicit example of how macroscopic dissipation and self-organization can result from an underlying microscopically reversible dynamics.

cond-mat.stat-mech↗