NEST

Network Science and Technology Center

The Network Science and Technology (NEST) Center is focused on the fundamental research and engineering of natural and technological networks, ranging from social and cognitive networks to computer networks. The fundamental understanding of network structures and dynamical processes arising in them combined with the novel designs of protocols for communication and algorithms for applications will enable experts in the fields ranging from sociology, to biology, medicine, physics, computer science and engineering and transportation engineering to apply the results of the center research in their specific disciplines.

NEST research is focused on studying fundamental properties of networks, the processes underlying their evolution and the paradigms for network engineering to enhance their efficiency, reliability, robustness and other desirable properties. Research on natural networks, such as social and cognitive networks in which people interact over variety of means, focuses on cognitive models of net-centric interactions, models and algorithms of community creation and evolution, impact of mobility on network formation, dependencies between social, information and communication networks and spread of opinions and ideologies among network nodes. Research on technological networks, such as computer, transportation and energy distribution networks, focuses on their optimal design from the point of view of flow maximization, fault tolerance to failure, graceful degradation in case of partial damage, etc. In communication networks, NEST develops and studies network protocols and algorithms, especially for wireless and sensor networks and studies system issues in interoperability of communication networks with computer systems. NEST actively transitions the developed protocols and algorithms to industrial practice and commercialization.

NEST partners with universities, national laboratories and industry in large scientific programs targeting interdisciplinary research. NEST is the primary member of the Social Cognitive Network Academic Research Center, a part of Network Science Collaborative Technology Alliance, as well as member of the International Technology Alliance, both funded by collaborative agreements with ARL.

Research Projects

2009–OngoingSocial/Cognitive Networks Academic Research Center, ARL/CTA

PI: Szymanski, Boleslaw

Website: scnarc.rpi.edu

The ARL Social Cognitive Network Academic Research Center (SCNARC) has been created and funded as a part of the US Army Network Science Collaborative Technology Alliance together with three other centers focusing on different kind of networks. The principal member of the Center is Rensselaer Polytechnic Institute while the remaining members are CUNY, IBM TJ Watson Research Laboratory, and Northeastern University. The Center includes also collaborators from the Army Research Laboratory, Indiana University, University of Maryland, MIT, Northwestern University, University of Notre Dame and NYU. The Center will collaborate closely with other centers of the Network Science CTA. Rapid growth of web-based social networks has redefined social interactions. Increasing popular, technology-based social networks do not require personal, direct contact but at the same time they provide rich traces of data about their activities. These kinds of social networks and the behaviors that govern their dynamics and evolution are the subject of our research.

2011–OngoingImplementation of Robust TCP for Airborne Networks, Air Force

PI: Kar, Koushik

This project involves implementation of a Loss-Tolerant Transport Protocol (LT-TCP) on a Linux platform. The project also includes evaluation of LT-TCP (with respect to TCP) in terms of goodput and latency under a wide range of loss rate conditions, and possible redesign and fine-tuning of the LT-TCP protocol as necessary. This work is being conducted in close cooperation with MIT Lincoln Laboratory.

2010–OngoingMetpetDB: A Database for Metamorphic Geochemistry, NSF

PI: Spear, Frank

MetPetDB is a database for metamorphic petrology that is being designed and built by a global community of metamorphic petrologists in collaboration with computer scientists at Rensselaer Polytechnic Institute as part of the National Cyberinfrastructure Initiative and supported by the National Science Foundation. This project supports the development, implementation and population of MetPetDB with the purpose of, archiving published data, storing new data for ready access to researchers and students, facilitating the gathering of information for researchers beginning new projects, providing a search mechanism for data relating to anywhere on the globe, providing a platform for collaborative studies among researchers, and serving as a portal for students beginning their studies of metamorphic geology.

2009–OngoingCommunity Stability and Social Engineering in Large-Scale Social Networks: Employing Individual-Based Models for Opinion Dynamics to Detect and Destabilize Communities, ONR

PI: Szymanski, Boleslaw

We study real data projected out from empirical or info-social network, to develop methodologies and techniques for extracting relevant information and behavioral patterns from the underlying social network. With the availability and accessibility of vast amount of data in recent years, our project advances our fundamental knowledge, from a network science viewpoint, and creates methodologies and techniques that could be applied to address strategic and urgent needs aligned with national priorities for network research. Using a novel combination of stochastic agent-based models for opinion formation and massive real-life data, we are developing models with predictive power for large-scale social networks.

2009–OngoingOptimizing Robustness of Large-Scale Information and Infrastructure Networks, DTRA

PI: Szymanski, Boleslaw

We study vulnerability and recovery in complex networks with the aim to develop methods and design a prototype system for optimizing robustness in such networks. We are developing generic analytic, numerical, and simulation methods to model and analyze, ``what if?" disruption scenarios and encapsulate such methods in a prototype system for such modeling and analysis. We also investigate methods for reallocation of resources for the surviving part of the network to kept it operational and efficient. We investigate flow in resistor networks and the resulting load landscapes that provide a fundamental model for transport, based on local routing and conservation laws. This basic model and the resulting load distribution and load landscape in a given network will be used to identify the nodes with the extreme loads in real-life empirical networks, and in turn, to optimize transport.

2009–OngoingCitizens Science: Enabling Computational Probabilistic Methods for Organism's Transcriptional Regulatory Network Using Voluntary Computing Platforms, NSF

PI: Szymanski, Boleslaw

We aim at demonstrating that BOINC can be an efficient platform for a wide class of applications employing a probabilistic inference based on Gibbs sampling. Specifically, we are developing an architecture by which Gibbs sampling applications can exploitBOINC and we are also developing the corresponding Gibbs sampling interface to BOINC, Using this interface, we are converting the Phylogenetic Gibbs Centroid Sampler to volunteering computing. The Phylogenetic Gibbs Centroid Sampler is the most powerful version of a 16-year-established molecular biology algorithm that yet needs a boost to solve gene regulatory problems with genome-scale sequence data sets.

2006–OngoingComplexity Management of Data Infrastructure, ARL/ITA

PI: Szymanski, Boleslaw

The project investigates sensor networks that consist of ensembles of low-cost, unreliable devices that collect measurements using a variety of sensing modalities and perform elementary processing on this raw data such as various aggregation operations by wirelessly transmitting the intermediate results over multiple hops on the way to their final destination to base-stations (static or mobile) for collection, storage and further analysis. This in-network processing is necessary due to the limited resources of the sensor devices (e.g. energy, bandwidth, processing capacity, etc.), as well as due to the unreliable communication medium that would make it prohibitively expensive to collect-and-transmit the typically very large set of initial, sampled data. The collective application of this distributed processing provides a composite service that can be used as higher-level intelligence for decision-making; it is termed sensor service composition. The focus of the project is on dynamic sensor service composition that not only optimizes the cost of the service but also ensures that the composition complies with the security policies.

2006–2010SEI(AST): A Dynamic Grid for Astroinformatics: Data-Driven Discovery of the Milky Way Origin and Evolution from the Sloan Digital Sky Survey, NSF

PI: Magdon-Ismail, Malik

This project marks a collaboration between Prof. Newberg of physics and Profs. Magdon-Ismail, Varela and Szymanski of Computer Science. The goal was to develop a parallel likelihood maximization technique which could automatically infer structure in the universe, in particular tidal galaxy streams. The project ultimately managed to leverage over 50,000 internet CPUs volunteered by clients across the globe over the BOINC infrastructure to deliver a computing power greater than most of the worlds supercomputers in a model of computing we call "Citizen's Science". The result was efficient discovery of Milky Way structure. We are now using this technology to trace back into the history of the universe.

2005–OngoingLocal Information Based Distributed Optimization of Resources in Large-Scale Adhoc and Sensor Networks, NSF/CAREER

PI: Kar, Koushik

This project focuses on developing fundamental theory and algorithms for bandwidth and energy optimization in large-scale adhoc and sensor networks, using tools and techniques from convex, combinatorial and stochastic optimization, statistical modeling and approximation techniques. A key focus of this project is on the development of resource optimization algorithms that can be executed based only on local information about the topology and state of the network, and yet guarantees a performance that closely approximates the global or local optimum. Specifically, the project focuses on key networking functions like scheduling, medium/spectrum access control, channel assignment and power allocation in wireless networks, and wakeup management and event detection in sensor networks, posing them as bandwidth and/or energy optimization questions.

2005–OngoingMiddleware and Programming Technology for Grid Computing, NSF/CAREER

PI: Varela, Carlos

2003–2010ITR: Study of Dynamically Evolving Social Groups in Communication Networks, NSF

PI: Goldberg, Mark

2001–OngoingScalable Network Performance Evaluation, CISCO

PI: Szymanski, Boleslaw

The project supports development of on-line network simulation as a tool for network performance evaluation. The goal is to enable large-scale network simulation by representing network domains at the packet level, while the interchanges between domains as packet flows. Using fixed point solution to such hierarchical simulation makes the approach scalable across wide range of networks.

2000–OngoingMetacomputing: Nomadic and Parallel Computation Over the Internet, IBM, SUR

PI: Szymanski, Boleslaw

The project focuses on parallel computations over the Internet, including volunteer computing. Two large application has been developed, one called MilkyWay@home that focuses on searching for streams of stars from Milky Way neighboring galaxies into Milky Way by its gravitational pull. The other application, DNA@home focuses on transcriptional regulatory network using voluntary computing platforms. In both cases, we are using BOINC infrastructure and server to dispatch work to users and collected results. The grant provided initial impetus for the computational work and is now supplemented by NSF support for relevant science research.

People

Staff
Affiliated Faculty
Postdoctoral Researchers