Nadya T. Bliss

Dynamic distributed dimensional data model (D4M) database and computation system

A crucial element of large web companies is their ability to collect and analyze massive amounts of data. Tuple store databases are a key enabling technology employed by many of these companies (e.g., Google Big Table and Amazon Dynamo). Tuple stores are highly scalable and run on commodity clusters, but lack interfaces to support efficient development of mathematically based analytics. D4M (Dynamic Distributed Dimensional Data Model) has been developed to provide a mathematically rich interface to tuple stores (and structured query language “SQL” databases). D4M allows linear algebra to be readily applied to databases. Using D4M, it is possible to create composable analytics with significantly less effort than using traditional approaches. This work describes the D4M technology and its application and performance.

Circuit-Switched Memory Access in Photonic Interconnection Networks for High-Performance Embedded Computing

As advancements in CMOS technology trend toward ever increasing core counts in chip multiprocessors for high-performance embedded computing, the discrepancy between on- and off-chip communication bandwidth continues to widen due to the power and spatial constraints of electronic off-chip signaling. Silicon photonics-based communication offers many advantages over electronics for network-on-chip design, namely power consumption that is effectively agnostic to distance traveled at the chip- and board-scale, even across chip boundaries. In this work we develop a design for a photonic network-on-chip with integrated DRAM I/O interfaces and compare its performance to similar electronic solutions using a detailed network-on-chip simulation. When used in a circuit-switched network, silicon nanophotonic switches offer higher bandwidth density and low power transmission, adding up to over 10x better performance and 3-5x lower power over the baseline for projective transform, matrix multiply, and Fast Fourier Transform (FFT), all key algorithms in embedded real-time signal and image processing.

Time-division-multiplexed arbitration in silicon nanophotonic networks-on-chip for high-performance chip multiprocessors

As the computational performance of microprocessors continues to grow through the integration of an increasing number of processing cores on a single die, the interconnection network has become the central subsystem for providing the communications infrastructure among the on-chip cores as well as to off-chip memory. Silicon nanophotonics as an interconnect technology offers several promising benefits for future networks-on-chip, including low end-to-end transmission energy and high bandwidth density of waveguides using wavelength division multiplexing. In this work, we propose the use of time-division-multiplexed distributed arbitration in a photonic mesh network composed of silicon micro-ring resonator based photonic switches, which provides round-robin fairness to setting up photonic circuit paths. Our design sustains over 10x more bandwidth and uses less power than the compared network designs. We also observe a 2x improvement in performance for memory-centric application traces using the MORE modeling system.

Toward signal processing theory for graphs and non-Euclidean data

Graphs are canonical examples of high-dimensional non-Euclidean data sets, and are emerging as a common data structure in many fields. While there are many algorithms to analyze such data, a signal processing theory for evaluating these techniques akin to detection and estimation in the classical Euclidean setting remains to be developed. In this paper we show the conceptual advantages gained by formulating graph analysis problems in a signal processing framework by way of a practical example: detection of a subgraph embedded in a background graph. We describe an approach based on detection theory and provide empirical results indicating that the test statistic proposed has reasonable power to detect dense subgraphs in large random graphs.

A scalable signal processing architecture for massive graph analysis

In many applications, it is convenient to represent data as a graph, and often these datasets will be quite large. This paper presents an architecture for analyzing massive graphs, with a focus on signal processing applications such as modeling, filtering, and signal detection. We describe the architecture, which covers the entire processing chain, from data storage to graph construction to graph analysis and subgraph detection. The data are stored in a new format that allows easy extraction of graphs representing any relationship existing in the data. The principal analysis algorithm is the partial eigendecomposition of the modularity matrix, whose running time is discussed. A large document dataset is analyzed, and we present subgraphs that stand out in the principal eigenspace of the time-varying graphs, including behavior we regard as clutter as well as small, tightly-connected clusters that emerge over time.

Matched filtering for subgraph detection in dynamic networks

Graphs are high-dimensional, non-Euclidean data, whose utility spans a wide variety of disciplines. While their non-Euclidean nature complicates the application of traditional signal processing paradigms, it is desirable to seek an analogous detection framework. In this paper we present a matched filtering method for graph sequences, extending to a dynamic setting a previous method for the detection of anomalously dense subgraphs in a large background. In simulation, we show that this temporal integration technique enables the detection of weak subgraph anomalies than are not detectable in the static case. We also demonstrate background/foreground separation using a real background graph based on a computer network.

Anomalous subgraph detection via Sparse Principal Component Analysis

Network datasets have become ubiquitous in many fields of study in recent years. In this paper we investigate a problem with applicability to a wide variety of domains — detecting small, anomalous subgraphs in a background graph. We characterize the anomaly in a subgraph via the well-known notion of network modularity, and we show that the optimization problem formulation resulting from our setup is very similar to a recently introduced technique in statistics called Sparse Principal Component Analysis (Sparse PCA), which is an extension of the classical PCA algorithm. The exact version of our problem formulation is a hard combinatorial optimization problem, so we consider a recently introduced semidefinite programming relaxation of the Sparse PCA problem. We show via results on simulated data that the technique is very promising.

pMATLAB: Parallel MATLAB Library for Signal Processing Applications

MATLAB® is one of the most commonly used languages for scientific computing with approximately one million users worldwide. At MIT Lincoln Laboratory, MATLAB is used by technical staff to develop sensor processing algorithms. MATLABs popularity is based on availability of high-level abstractions leading to reduced code development time. Due to the compute intensive nature of scientific computing, these applications often require long running times and would benefit greatly from increased performance offered by parallel computing. pMatlab (www.ll.mit.edu/pMatlab) implements partitioned global address space (PGAS) support via standard operator overloading techniques. The core data structures in pMatlab are distributed arrays and maps, which simplify parallel programming by removing the need for explicit message passing. This paper presents the pMatlab design and results for the HPC Challenge benchmark suite. Additionally, two case studies of pMatlab use are described.

Toward matched filter optimization for subgraph detection in dynamic networks

This paper outlines techniques for optimization of filter coefficients in a spectral framework for anomalous subgraph detection. Restricting the scope to the detection of a known signal in i.i.d. noise, the optimal coefficients for maximizing the signals power are shown to be found via a rank-1 tensor approximation of the subgraphs dynamic topology. While this technique optimizes our power metric, a filter based on average degree is shown in simulation to work nearly as well in terms of power maximization and detection performance, and better separates the signal from the noise in the eigenspace.

High-Productivity Software Development with pMatlab

In this paper, we explore the ease of tackling a communication-intensive parallel computing task - namely, the 2D fast Fourier transform (FFT). We start with a simple serial Matlab code, explore in detail a ID parallel FFT, and illustrate how it can be extended to multidimensional FFTs.