Neutron transport simulation plays a central role in nuclear engineering, yet its inverse problems—such as shielding design and sensitivity analysis—remain extremely challenging. The governing transport equation shares a common mathematical foundation with the volume rendering equation used for participating media in computer graphics. Recent advances in differentiable rendering have demonstrated that inverse problems—such as recovering scene properties from observations—can be efficiently solved through gradient-based optimization, where the gradients are estimated through Monte Carlo methods. Inspired by these developments, we extend differentiable rendering techniques to neutron transport problems, enabling inverse design and sensitivity analysis in nuclear systems. To adapt these methods to the neutron transport problem, we generalize the automatic differentiation of Monte Carlo light transport simulation to multi-energy neutron transport; propose a new simplified way to handle moving geometric discontinuities; and propose new estimators for derivatives of volume integrals. We demonstrate the effectiveness of the proposed framework on several model inverse problems, including gradient-based optimization of multi-group neutron moderation, sensitivity analysis of material properties, and diagnostic sensor shape optimization. These results show that differentiable Monte Carlo neutron transport can support optimization and analysis tasks, opening new opportunities for inverse design in nuclear shielding and related applications.

Poisson Surface Reconstruction is a widely-used algorithm for reconstructing a surface from an oriented point cloud. To facilitate applications where only partial surface information is available, or scanning is performed sequentially, a recent line of work proposes to incorporate uncertainty into the reconstructed surface via Gaussian process models. The resulting algorithms first perform Gaussian process interpolation, then solve a set of volumetric partial differential equations globally in space, resulting in a computationally expensive two-stage procedure. In this work, we apply recently-developed techniques from geometric Gaussian processes to combine interpolation and surface reconstruction into a single stage, requiring only one linear solve per sample. The resulting reconstructed surface samples can be queried locally in space, without the use of problem-dependent volumetric meshes or grids. These capabilities enable one to (a) perform probabilistic collision detection locally around the region of interest, (b) perform ray casting without evaluating points not on the ray’s trajectory, and (c) perform next-view planning on a per-ray basis. They also do not requiring one to approximate kernel matrix inverses with diagonal matrices as part of intermediate computations, unlike prior methods. Results show that our approach provides a cleaner, more-principled, and more-flexible stochastic surface reconstruction pipeline.

Backward error analysis offers a method for assessing the quality of numerical programs in the presence of floating-point rounding errors. However, techniques from the numerical analysis literature for quantifying backward error require substantial human effort, and there are currently no tools or automated methods for statically deriving sound backward error bounds. To address this gap, we propose Bean , a typed first-order programming language designed to express quantitative bounds on backward error. Bean ’s type system combines a graded coeffect system with strict linearity to soundly track the flow of backward error through programs. We prove the soundness of our system using a novel categorical semantics, where every Bean program denotes a triple of related transformations that together satisfy a backward error guarantee. To illustrate Bean ’s potential as a practical tool for automated backward error analysis, we implement a variety of standard algorithms from numerical linear algebra in Bean , establishing fine-grained backward error bounds via typing in a compositional style. We also develop a prototype implementation of Bean that infers backward error bounds automatically. Our evaluation shows that these inferred bounds match worst-case theoretical relative backward error bounds from the literature, underscoring Bean ’s utility in validating a key property of numerical programs: numerical stability .

Community is a universal structure in various complex networks, and community detection is a fundamental task for network analysis. With the rapid growth of network scale, networks are massive, changing rapidly, and could naturally be modeled as graph streams. Due to the limited memory and access constraint in graph streams, existing non-streaming community detection methods are no longer applicable. This raises an emerging need for online approaches. In this work, we consider the problem of uncovering the local community containing a few query nodes in graph streams, termed streaming local community detection. This new problem raised recently is more challenging for community detection, and only a few works address this online setting. Correspondingly, we design an online single-pass streaming local community detection approach. Inspired by the local property of communities, our method samples the local structure around the query nodes in graph streams and extracts the target community on the sampled subgraph using our proposed metric called approximate conductance. Comprehensive experiments show that our method remarkably outperforms the streaming baseline on both effectiveness and efficiency, and even achieves similar accuracy compared to the state-of-the-art non-streaming local community detection methods that use static and complete graphs.

We consider a task of surveillance-evading path-planning in a continuous setting. An Evader strives to escape from a 2D domain while minimizing the risk of detection (and immediate capture). The probability of detection is path-dependent and determined by the spatially inhomogeneous surveillance intensity, which is fixed but a priori unknown and gradually learned in the multi-episodic setting. We introduce a Bayesian reinforcement learning algorithm that relies on a Gaussian Process regression (to model the surveillance intensity function based on the information from prior episodes), numerical methods for Hamilton-Jacobi PDEs (to plan the best continuous trajectories based on the current model), and Confidence Bounds (to balance the exploration vs exploitation). We use numerical experiments and regret metrics to highlight the significant advantages of our approach compared to traditional graph-based algorithms of reinforcement learning.

The LAProof library provides formal machine-checked proofs of the accuracy of basic linear algebra operations: inner product using conventional multiply and add, inner product using fused multiply-add, scaled matrix-vector and matrix-matrix multiplication, and scaled vector and matrix addition. These proofs can connect to concrete implementations of low-level basic linear algebra subprograms; as a proof of concept we present a machine-checked correctness proof of a C function implementing sparse matrix-vector multiplication using the compressed sparse row format. Our accuracy proofs are backward error bounds and mixed backward-forward error bounds that account for underflow, proved subject to no assumptions except a low-level formal model of IEEE-754 arithmetic. We treat low-order error terms concretely, not approximating as $\mathcal{O}\left( {{u^2}} \right)$.

Variational Gaussian processes (GPs) approximate exact GP inference by using a small set of inducing points to form a sparse approximation of the true posterior, with the fidelity of the model increasing with additional inducing points. Although the approximation error in principle can be reduced through the use of more inducing points, this leads to scaling optimization challenges and computational complexity. To achieve scalability, inducing point methods typically introduce conditional independencies and then approximations to the training and test conditional distributions. In this paper, we consider an alternative approach to modifying the training and test conditionals, in which we make them more flexible. In particular, we investigate decoupling the parametric form of the predictive mean and covariance in the conditionals, and learn independent parameters for predictive mean and covariance. We derive new evidence lower bounds (ELBO) under these more flexible conditionals, and provide two concrete examples of applying the decoupled conditionals. Empirically, we find this additional flexibility leads to improved model performance on a variety of regression tasks and Bayesian optimization (BO) applications.

Across many data domains, co-occurrence statistics about the joint appearance of objects are powerfully informative. By transforming unsupervised learning problems into decompositions of co-occurrence statistics, spectral algorithms provide transparent and efficient algorithms for posterior inference such as latent topic analysis and community detection. As object vocabularies grow, however, it becomes rapidly more expensive to store and run inference algorithms on co-occurrence statistics. Rectifying co-occurrence, the key process to uphold model assumptions, becomes increasingly more vital in the presence of rare terms, but current techniques cannot scale to large vocabularies. We propose novel methods that simultaneously compress and rectify co-occurrence statistics, scaling gracefully with the size of vocabulary and the dimension of latent space. We also present new algorithms learning latent variables from the compressed statistics, and verify that our methods perform comparably to previous approaches on both textual and non-textual data.

Gaussian processes with derivative information are useful in many settings where derivative information is available, including numerous Bayesian optimization and regression tasks that arise in the natural sciences. Incorporating derivative observations, however, comes with a dominating $O(N^3D^3)$ computational cost when training on $N$ points in $D$ input dimensions. This is intractable for even moderately sized problems. While recent work has addressed this intractability in the low-$D$ setting, the high-$N$, high-$D$ setting is still unexplored and of great value, particularly as machine learning problems increasingly become high dimensional. In this paper, we introduce methods to achieve fully scalable Gaussian process regression with derivatives using variational inference. Analogous to the use of inducing values to sparsify the labels of a training set, we introduce the concept of inducing directional derivatives to sparsify the partial derivative information of the training set. This enables us to construct a variational posterior that incorporates derivative information but whose size depends neither on the full dataset size $N$ nor the full dimensionality $D$. We demonstrate the full scalability of our approach on a variety of tasks, ranging from a high dimensional Stellarator fusion regression task to training graph convolutional neural networks on PubMed using Bayesian optimization. Surprisingly, we additionally find that our approach can improve regression performance even in settings where only label data is available.

A fundamental problem on graph-structured data is that of quantifying similarity between graphs. Graph kernels are an established technique for such tasks; in particular, those based on random walks and return probabilities have proven to be effective in wide-ranging applications, from bioinformatics to social networks to computer vision. However, random walk kernels generally suffer from slowness and tottering, an effect which causes walks to overemphasize local graph topology, undercutting the importance of global structure. To correct for these issues, we recast return probability graph kernels under the more general framework of density of states — a framework which uses the lens of spectral analysis to uncover graph motifs and properties hidden within the interior of the spectrum — and use our interpretation to construct scalable, composite density of states based graph kernels which balance local and global information, leading to higher classification accuracies on benchmark datasets.

Spectral algorithms operate on matrices or tensors of word co-occurrence to learn latent topics. These approaches remove the dependence on the original documents and produce substantial gains in efficiency with provable inference, but at a cost: the models can no longer infer any information about individual documents. Thresholded Linear Inverse is developed to learn document-specific topic compositions, but its linear characteristics limit the inference quality without considering any prior information on topic distributions. We propose two novel estimation methods that respect previously unclear prior structures of spectral topic models. Experiments on a variety of synthetic to real collections demonstrate that our Prior-Aware Dual Decomposition outperforms the baseline method, whereas our Prior-Aware Manifold Iteration performs even better on short realistic data.

Rotation Averaging is a non-convex optimization problem that determines orientations of a collection of cameras from their images of a 3D scene. The problem has been studied using a variety of distances and robustifiers. The intrinsic (or geodesic) distance on SO(3) is geometrically meaningful; but while some extrinsic distance-based solvers admit (conditional) guarantees of correctness, no comparable results have been found under the intrinsic metric. In this paper, we study the spatial distribution of local minima. First, we do a novel empirical study to demonstrate sharp transitions in qualitative behavior: as problems become noisier, they transition from a single (easy-to-find) dominant minimum to a cost surface filled with minima. In the second part of this paper we derive a theoretical bound for when this transition occurs. This is an extension of the results of [24], which used local convexity as a proxy to study the difficulty of problem. By recognizing the underly- ing quotient manifold geometry of the problem we achieve an n-fold improvement over prior work. Incidentally, our analysis also extends the prior l2 work to general lp costs. Our results suggest using algebraic connectivity as an indicator of problem difficulty.

Gaussian processes (GPs) provide flexible distributions over functions, with inductive biases controlled by a kernel. However, in many applications Gaussian processes can struggle with even moderate input dimensionality. Learning a low dimensional projection can help alleviate this curse of dimensionality, but introduces many trainable hyperparameters, which can be cumbersome, especially in the small data regime. We use additive sums of kernels for GP regression, where each kernel operates on a different random projection of its inputs. Surprisingly, we find that as the number of random projections increases, the predictive performance of this approach quickly converges to the performance of a kernel operating on the original full dimensional inputs, over a wide range of data sets, even if we are projecting into a single dimension. As a consequence, many problems can remarkably be reduced to one dimensional input spaces, without learning a transformation. We prove this convergence and its rate, and additionally propose a deterministic approach that converges more quickly than purely random projections. Moreover, we demonstrate our approach can achieve faster inference and improved predictive accuracy for high-dimensional inputs compared to kernels in the original input space.

Bayesian optimization (BO) is a class of sample-efficient global optimization methods, where a probabilistic model conditioned on previous observations is used to determine future evaluations via the optimization of an acquisition function. Most acquisition functions are myopic, meaning that they only consider the impact of the next function evaluation. Non-myopic acquisition functions consider the impact of the next h function evaluations and are typically computed through rollout, in which h steps of BO are simulated. These rollout acquisition functions are defined as h-dimensional integrals, and are expensive to compute and optimize. We show that a combination of quasi-Monte Carlo, common random numbers, and control variates significantly reduce the computational burden of rollout. We then formulate a policy-search based approach that removes the need to optimize the rollout acquisition function. Finally, we discuss the qualitative behavior of rollout policies in the setting of multi-modal objectives and model error.

Existing fiber scattering models in rendering are all based on tracing rays through fiber geometry, but for small fibers diffraction and interference are non-negligible, so relying on ray optics can result in appearance errors. This paper presents the first wave optics based fiber scattering model, introducing an azimuthal scattering function that comes from a full wave simulation. Solving Maxwell's equations for a straight fiber of constant cross section illuminated by a plane wave reduces to solving for a 3D electromagnetic field in a 2D domain, and our fiber scattering simulator solves this 2.5D problem efficiently using the boundary element method (BEM). From the resulting fields we compute extinction, absorption, and far-field scattering distributions, which we use to simulate shadowing and scattering by fibers in a path tracer. We validate our path tracer against the wave simulation and the simulation against a measurement of diffraction from a single textile fiber. Our results show that our approach can reproduce a wide range of fibers with different sizes, cross sections, and material properties, including textile fibers, animal fur, and human hair. The renderings include color effects, softening of sharp features, and strong forward scattering that are not predicted by traditional ray-based models, though the two approaches produce similar appearance for complex fiber assemblies under many conditions.

Moontae Lee, Sungjun Cho, David Bindel, David Mimno. Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP). 2019.

Spectral analysis connects graph structure to the eigenvalues and eigenvectors of associated matrices. Much of spectral graph theory descends directly from spectral geometry, the study of differentiable manifolds through the spectra of associated differential operators. But the translation from spectral geometry to spectral graph theory has largely focused on results involving only a few extreme eigenvalues and their associated eigenvalues. Unlike in geometry, the study of graphs through the overall distribution of eigenvalues --- the \em spectral density --- is largely limited to simple random graph models. The interior of the spectrum of real-world graphs remains largely unexplored, difficult to compute and to interpret. In this paper, we delve into the heart of spectral densities of real-world graphs. We borrow tools developed in condensed matter physics, and add novel adaptations to handle the spectral signatures of common graph motifs. The resulting methods are highly efficient, as we illustrate by computing spectral densities for graphs with over a billion edges on a single compute node. Beyond providing visually compelling fingerprints of graphs, we show how the estimation of spectral densities facilitates the computation of many common centrality measures, and use spectral densities to estimate meaningful information about graph structure that cannot be inferred from the extremal eigenpairs alone.

Inferring electromagnetic (EM) propagation characteristics within the marine atmospheric boundary layer (MABL) from data in real time is crucial for modern maritime navigation and communications. The propagation of EM waves is well modeled by a partial differential equation (PDE): a Helmholtz equation. A natural way to solve the MABL characterization inverse problem is to minimize what is observed and what is predicted by the PDE. However, this optimization is difficult because it has many local minima. We propose an alternative solution that relies on the properties of the PDE but does not involve solving the full forward model. Ducted environments result in an EM field which can be decomposed into a few propagating, trapped modes. These modes are a subset of the solutions to a Sturm-Liouville eigenvalue problem. We design a new objective function that measures the distance from the observations to a subspace spanned by these eigenvectors. The resulting optimization problem is much easier than the one that arises in the standard approach, and we show how to solve the associated nonlinear eigenvalue problem efficiently, leading to a real-time method.

Community detection is an important information mining task to uncover modular structures in large networks. For increasingly common large network datasets, global community detection is prohibitively expensive, and attention has shifted to methods that mine local communities, i.e., identifying all latent members of a particular community from a few labeled seed members. To address such semi-supervised mining task, we systematically develop a local spectral (LOSP) subspace-based community detection method, called LOSP. We define a family of LOSP subspaces based on Krylov subspaces, and seek a sparse indicator for the target community via an ℓ 1 norm minimization over the Krylov subspace. Variants of LOSP depend on type of random walks with different diffusion speeds, type of random walks, dimension of the LOSP subspace, and step of diffusions. The effectiveness of the proposed LOSP approach is theoretically analyzed based on Rayleigh quotients, and it is experimentally verified on a wide variety of real-world networks across social, production, and biological domains, as well as on an extensive set of synthetic LFR benchmark datasets.

Gaussian processes (GPs) with derivatives are useful in many applications, including Bayesian optimization, implicit surface reconstruction, and terrain reconstruction. Fitting a GP to function values and derivatives at $n$ points in $d$ dimensions requires linear solves and log determinants with an ${n(d+1) \times n(d+1)}$ positive definite matrix-- leading to prohibitive $\mathcal{O}(n^3d^3)$ computations for standard direct methods. We propose iterative solvers using fast $\mathcal{O}(nd)$ matrix-vector multiplications (MVMs), together with pivoted Cholesky preconditioning that cuts the iterations to convergence by several orders of magnitude, allowing for fast kernel learning and prediction. Our approaches, together with dimensionality reduction, allows us to scale Bayesian optimization with derivatives to high-dimensional problems and large evaluation budgets.

Despite advances in scalable models, the inference tools used for Gaussian processes (GPs) have yet to fully capitalize on developments in computing hardware. We present an efficient and general approach to GP inference based on Blackbox Matrix-Matrix multiplication (BBMM). BBMM inference uses a modified batched version of the conjugate gradients algorithm to derive all terms for training and inference in a single call. BBMM reduces the asymptotic complexity of exact GP inference from O(n^3) to O(n^2). Adapting this algorithm to scalable approximations and complex GP models simply requires a routine for efficient matrix-matrix multiplication with the kernel and its derivative. In addition, BBMM uses a specialized preconditioner to substantially speed up convergence. In experiments we show that BBMM effectively uses GPU hardware to dramatically accelerate both exact GP inference and scalable approximations. Additionally, we provide GPyTorch, a software platform for scalable GP inference via BBMM, built on PyTorch.

Large graphs arise in a number of contexts and understanding their structure and extracting information from them is an important research area. Early algorithms for mining communities have focused on global graph structure, and often run in time proportional to the size of the entire graph. As we explore networks with millions of vertices and find communities of size in the hundreds, it becomes important to shift our attention from macroscopic structure to microscopic structure in large networks. A growing body of work has been adopting local expansion methods in order to identify communities from a few exemplary seed members. In this article, we propose a novel approach for finding overlapping communities called L emon ( L ocal E xpansion via M inimum O ne N orm). Provided with a few known seeds , the algorithm finds the community by performing a local spectral diffusion. The core idea of L emon is to use short random walks to approximate an invariant subspace near a seed set, which we refer to as local spectra . Local spectra can be viewed as the low-dimensional embedding that captures the nodes’ closeness in the local network structure. We show that L emon ’s performance in detecting communities is competitive with state-of-the-art methods. Moreover, the running time scales with the size of the community rather than that of the entire graph. The algorithm is easy to implement and is highly parallelizable. We further provide theoretical analysis of the local spectral properties, bounding the measure of tightness of extracted community using the eigenvalues of graph Laplacian. We thoroughly evaluate our approach using both synthetic and real-world datasets across different domains, and analyze the empirical variations when applying our method to inherently different networks in practice. In addition, the heuristics on how the seed set quality and quantity would affect the performance are provided.

For applications as varied as Bayesian neural networks, determinantal point processes, elliptical graphical models, and kernel learning for Gaussian processes (GPs), one must compute a log determinant of an n by n positive definite matrix, and its derivatives---leading to prohibitive O(n^3) computations. We propose novel O(n) approaches to estimating these quantities from only fast matrix vector multiplications (MVMs). These stochastic approximations are based on Chebyshev, Lanczos, and surrogate models, and converge quickly even for kernel matrices that have challenging spectra. We leverage these approximations to develop a scalable Gaussian process approach to kernel learning. We find that Lanczos is generally superior to Chebyshev for kernel learning, and that a surrogate approach can be highly efficient and accurate with popular kernels.

Based on the definition of local spectral subspace, we propose a novel approach called LOSP for local overlapping community detection. Using the power method for a few steps, LOSP finds an approximate invariant subspace, which depicts the embedding of the local neighborhood structure around the seeds of interest. LOSP then identifies the local community expanded from the given seeds by seeking a sparse indicator vector in the subspace where the seeds are in its support. We provide a systematic investigation on LOSP, and thoroughly evaluate it on large real world networks across multiple domains. With the prior information of very few seed members, LOSP can detect the remaining members of a target community with high accuracy. Experiments demonstrate that LOSP outperforms the Heat Kernel and PageRank diffusions. Using LOSP as a subroutine, we further address the problem of multiple membership identification, which aims to find all the communities a single vertex belongs to. High F1 scores are achieved in detecting multiple local communities with respect to arbitrary single seed for various large real world networks.

Personalized PageRank is a standard tool for finding vertices in a graph that are most relevant to a query or user. To personalize PageRank, one adjusts node weights or edge weights that determine teleport probabilities and transition probabilities in a random surfer model. There are many fast methods to approximate PageRank when the node weights are personalized; however, personalization based on edge weights has been an open problem since the dawn of personalized PageRank over a decade ago. In this paper, we describe the first fast algorithm for computing PageRank on general graphs when the edge weights are personalized. Our method, which is based on model reduction, outperforms existing methods by nearly five orders of magnitude. This huge performance gain over previous work allows us --- for the very first time --- to solve learning-to-rank problems for edge weight personalization at interactive speeds, a goal that had not previously been achievable for this class of problems.

Configuring large distributed computations is a challenging task. Efficiently executing distributed computations requires configuration tuning based on careful examination of application and hardware properties. Considering the large number of parameters and impracticality of using trial and error in a production environment, programmers tend to make these decisions based on their experience and rules of thumb. Such configurations can lead to underutilized and costly clusters, and missed deadlines.

Spectral inference provides fast algorithms and provable optimality for latent topic analysis. But for real data these algorithms require additional ad-hoc heuristics, and even then often produce unusable results. We explain this poor performance by casting the problem of topic inference in the framework of Joint Stochastic Matrix Factorization (JSMF) and showing that previous methods violate the theoretical conditions necessary for a good solution to exist. We then propose a novel rectification method that learns high quality topics and their interactions even on small, noisy data. This method achieves results comparable to probabilistic techniques in several domains while maintaining scalability and provable optimality.

Large graphs arise in a number of contexts and understanding their structure and extracting information from them is an important research area. Early algorithms on mining communities have focused on the global structure, and often run in time functional to the size of the entire graph. Nowadays, as we often explore networks with billions of vertices and find communities of size hundreds, it is crucial to shift our attention from macroscopic structure to microscopic structure when dealing with large networks. A growing body of work has been adopting local expansion methods in order to identify the community from a few exemplary seed members. %Very few approaches can systematically demonstrate both high efficiency and effectiveness that significantly stands out amongst the divergent approaches in finding communities.

Scaling iterative graph processing applications to large graphs is an important problem. Performance is critical, as data scientists need to execute graph programs many times with varying parameters. The need for a high-level, high-performance programming model has inspired much research on graph programming frameworks. In this paper, we show that the important class of computationally light graph applications - applications that perform little computation per vertex - has severe scalability problems across multiple cores as these applications hit an early "memory wall" that limits their speedup. We propose a novel block-oriented computation model, in which computation is iterated locally over blocks of highly connected nodes, significantly improving the amount of computation per cache miss. Following this model, we describe the design and implementation of a block-aware graph processing runtime that keeps the familiar vertex-centric programming paradigm while reaping the benefits of block-oriented execution. Our experiments show that block-oriented execution significantly improves the performance of our framework for several graph applications.

Scientists are currently evaluating the cloud as a new platform. Many important scientific applications, however, perform poorly in the cloud. These applications proceed in highly parallel discrete time-steps or "ticks," using logical synchronization barriers at tick boundaries. We observe that network jitter in the cloud can severely increase the time required for communication in these applications, significantly increasing overall running time.

A long-standing line of work in economic theory has studied models by which a group of people in a social network, each holding a numerical opinion, can arrive at a shared opinion through repeated averaging with their neighbors in the network. Motivated by the observation that consensus is rarely reached in real opinion dynamics, we study a related sociological model in which individuals' intrinsic beliefs counterbalance the averaging process and yield a diversity of opinions. By interpreting the repeated averaging as best-response dynamics in an underlying game with natural payoffs, and the limit of the process as an equilibrium, we are able to study the cost of disagreement in these models relative to a social optimum. We provide a tight bound on the cost at equilibrium relative to the optimum, our analysis draws a connection between these agreement models and extremal problems for generalized eigenvalues. We also consider a natural network design problem in this setting, where adding links to the underlying network can reduce the cost of disagreement at equilibrium.

Internet fault diagnosis is extremely important for end users, overlay network service providers (like Akamai [1]) and even Internet service providers (ISPs). However, because link-level properties cannot be uniquely determined from end-to-end measurements, the accuracy of existing statistical diagnosis approaches is subject to uncertainty from statistical assumptions about the network. In this paper, we propose a novel Least-biased End-to-end Network Diagnosis (in short, LEND) system for inferring link-level properties like loss rate. We define a minimal identifiable link sequence (MILS) as a link sequence of minimal length whose properties can be uniquely identified from end-to-end measurements. We also design efficient algorithms to find all the MILSes and infer their loss rates for diagnosis. Our LEND system works for any network topology and for both directed and undirected properties, and incrementally adapts to network topology and property changes. It gives highly accurate estimates of the loss rates of MILSes, as indicated by both extensive simulations and Internet experiments. Furthermore, we demonstrate that such diagnosis can be achieved with fine granularity and in near real-time even for reasonably large overlay networks. Finally, LEND can supplement existing statistical inference approaches and provide smooth tradeoff between diagnosis accuracy and granularity.

Overlay network monitoring enables distributed Internet applications to detect and recover from path outages and periods of degraded performance within seconds. For an overlay network with n end hosts, existing systems either require O(n2) measurements, and thus lack scalability, or can only estimate the latency but not congestion or failures. Our earlier extended abstract [1] briefly proposes an algebraic approach that selectively monitors k linearly independent paths that can fully describe all the O(n2) paths. The loss rates and latency of these k paths can be used to estimate the loss rates and latency of all other paths. Our scheme only assumes knowledge of the underlying IP topology, with links dynamically varying between lossy and normal.In this paper, we improve, implement and extensively evaluate such a monitoring system. We further make the following contributions: i) scalability analysis indicating that for reasonably large n (e.g., 100), the growth of k is bounded as O(n log n), ii) efficient adaptation algorithms for topology changes, such as the addition or removal of end hosts and routing changes, iii) measurement load balancing schemes, and iv) topology measurement error handling. Both simulation and Internet experiments demonstrate we obtain highly accurate path loss rate estimation while adapting to topology changes within seconds and handling topology errors.

Overlay network monitoring enables distributed Internet applications to detect and recover from path outages and periods of degraded performance within seconds. For an overlay network with n end hosts, existing systems either require O(n2) measurements, and thus lack scalability, or can only estimate the latency but not congestion or failures. Unlike other network tomography systems, we characterize end-to-end losses (this extends to any additive metrics, including latency) rather than individual link losses. We find a minimal basis set of k linearly independent paths that can fully describe all the O(n,2) paths. We selectively monitor and measure the loss rates of these paths, then apply them to estimate the loss rates of all other paths. By extensively studying synthetic and real topologies, we find that for reasonably large n (e.g., 100), k is only in the range of O(n log n). This is explained by the moderately hierarchical nature of Internet routine.Our scheme only assumes the knowledge of underlying IP topology, and any link can become lossy or return to normal. In addition, our technique is tolerant to topology measurement inaccuracies, and is adaptive to topology changes.

We consider the efficient and accurate computation of Givens rotations. When f and g are positive real numbers, this simply amounts to computing the values of c = f /√ f 2 + g 2 , s = g /√ f 2 + g 2 , and r = √ f 2 + g 2 . This apparently trivial computation merits closer consideration for the following three reasons. First, while the definitions of c , s and r seem obvious in the case of two nonnegative arguments f and g , there is enough freedom of choice when one or more of f and g are negative, zero or complex that LAPACK auxiliary routines SLARTG, CLARTG, SLARGV and CLARGV can compute rather different values of c , s and r for mathematically identical values of f and g . To eliminate this unnecessary ambiguity, the BLAS Technical Forum chose a single consistent definition of Givens rotations that we will justify here. Second, computing accurate values of c , s and r as efficiently as possible and reliably despite over/underflow is surprisingly complicated. For complex Givens rotations, the most efficient formulas require only one real square root and one real divide (as well as several much cheaper additions and multiplications), but a reliable implementation using only working precision has a number of cases. On a Sun Ultra-10, the new implementation is slightly faster than the previous LAPACK implementation in the most common case, and 2.7 to 4.6 times faster than the corresponding vendor, reference or ATLAS routines. It is also more reliable; all previous codes occasionally suffer from large inaccuracies due to over/underflow. For real Givens rotations, there are also improvements in speed and accuracy, though not as striking. Third, the design process that led to this reliable implementation is quite systematic, and could be applied to the design of similarly reliable subroutines.

OceanStore is a utility infrastructure designed to span the globe and provide continuous access to persistent information. Since this infrastructure is comprised of untrusted servers, data is protected through redundancy and cryptographic techniques. To improve performance, data is allowed to be cached anywhere, anytime. Additionally, monitoring of usage patterns allows adaptation to regional outages and denial of service attacks; monitoring also enhances performance through pro-active movement of data. A prototype implementation is currently under development.