The acceleration of gradient-based optimization methods is a subject of significant practical and theoretical importance, particularly within machine learning applications. While much attention has been directed towards optimizing within Euclidean space, the need to optimize over spaces of probability measures in machine learning motivates the exploration of accelerated gradient methods in this context, too. To this end, we introduce a Hamiltonian-flow approach analogous to momentum-based approaches in Euclidean space. We demonstrate that, in the continuous-time setting, algorithms based on this approach can achieve convergence rates of arbitrarily high order. We complement our findings with numerical examples.

We analyze the sample complexity of single-loop quadratic penalty and augmented Lagrangian algorithms for solving nonconvex optimization problems with functional equality constraints. We consider three cases, in all of which the objective is stochastic, that is, an expectation over an unknown distribution that is accessed by sampling. The nature of the equality constraints differs among the three cases: deterministic and linear in the first case, deterministic and nonlinear in the second case, and stochastic and nonlinear in the third case. Variance reduction techniques are used to improve the complexity. To find a point that satisfies $\varepsilon$-approximate first-order conditions, we require $\widetilde{O}(\varepsilon^{-3})$ complexity in the first case, $\widetilde{O}(\varepsilon^{-4})$ in the second case, and $\widetilde{O}(\varepsilon^{-5})$ in the third case. For the first and third cases, they are the first algorithms of “single loop” type that also use $O(1)$ samples at each iteration and still achieve the best-known complexity guarantees.

We focus on constrained, $L$-smooth, potentially stochastic and nonconvex-nonconcave min-max problems either satisfying $\rho$-cohypomonotonicity or admitting a solution to the $\rho$-weakly Minty Variational Inequality (MVI), where larger values of the parameter $\rho>0$ correspond to a greater degree of nonconvexity. These problem classes include examples in two player reinforcement learning, interaction dominant min-max problems, and certain synthetic test problems on which classical min-max algorithms fail. It has been conjectured that first-order methods can tolerate a value of $\rho$ no larger than $\frac{1}{L}$, but existing results in the literature have stagnated at the tighter requirement $\rho < \frac{1}{2L}$. With a simple argument, we obtain optimal or best-known complexity guarantees with cohypomonotonicity or weak MVI conditions for $\rho < \frac{1}{L}$. First main insight for the improvements in the convergence analyses is to harness the recently proposed conic nonexpansiveness property of operators. Second, we provide a refined analysis for inexact Halpern iteration that relaxes the required inexactness level to improve some state-of-the-art complexity results even for constrained stochastic convex-concave min-max problems. Third, we analyze a stochastic inexact Krasnosel’skii-Mann iteration with a multilevel Monte Carlo estimator when the assumptions only hold with respect to a solution.

We leverage convex and bilevel optimization techniques to develop a general gradient-based parameter learning framework for neural-symbolic (NeSy) systems. We demonstrate our framework with NeuPSL, a state-of-the-art NeSy architecture. To achieve this, we propose a smooth primal and dual formulation of NeuPSL inference and show learning gradients are functions of the optimal dual variables. Additionally, we develop a dual block coordinate descent algorithm for the new formulation that naturally exploits warm-starts. This leads to over $100 \times$ learning runtime improvements over the current best NeuPSL inference method. Finally, we provide extensive empirical evaluations across $8$ datasets covering a range of tasks and demonstrate our learning framework achieves up to a $16$% point prediction performance improvement over alternative learning methods.

We revisit the problem of federated learning (FL) with private data from people who do not trust the server or other silos/clients. In this context, every silo (e.g. hospital) has data from several people (e.g. patients) and needs to protect the privacy of each person's data (e.g. health records), even if the server and/or other silos try to uncover this data. Inter-Silo Record-Level Differential Privacy (ISRL-DP) prevents each silo's data from being leaked, by requiring that silo $i$'s *communications* satisfy item-level differential privacy. Prior work (Lowy & Razaviyayn, 2023a) characterized the optimal excess risk bounds for ISRL-DP algorithms with *homogeneous* (i.i.d.) silo data and convex loss functions. However, two important questions were left open: 1) Can the same excess risk bounds be achieved with *heterogeneous* (non-i.i.d.) silo data? 2) Can the optimal risk bounds be achieved with *fewer communication rounds*? In this paper, we give positive answers to both questions. We provide novel ISRL-DP FL algorithms that achieve the optimal excess risk bounds in the presence of heterogeneous silo data. Moreover, our algorithms are more *communication-efficient* than the prior state-of-the-art. For smooth loss functions, our algorithm achieves the *optimal* excess risk bound and has *communication complexity that matches the non-private lower bound*. Additionally, our algorithms are more *computationally efficient* than the previous state-of-the-art.

We provide a simple and flexible framework for designing differentially private algorithms to find approximate stationary points of non-convex loss functions. Our framework is based on using a private approximate risk minimizer to "warm start" another private algorithm for finding stationary points. We use this framework to obtain improved, and sometimes optimal, rates for several classes of non-convex loss functions. First, we obtain improved rates for finding stationary points of smooth non-convex empirical loss functions. Second, we specialize to quasar-convex functions, which generalize star-convex functions and arise in learning dynamical systems and training some neural nets. We achieve the optimal rate for this class. Third, we give an optimal algorithm for finding stationary points of functions satisfying the Kurdyka-Lojasiewicz (KL) condition. For example, over-parameterized neural networks often satisfy this condition. Fourth, we provide new state-of-the-art rates for stationary points of non-convex population loss functions. Fifth, we obtain improved rates for non-convex generalized linear models. A modification of our algorithm achieves nearly the same rates for second-order stationary points of functions with Lipschitz Hessian, improving over the previous state-of-the-art for each of the above problems.

Nonconvex optimization is central in solving many machine learning problems, in which block-wise structure is commonly encountered. In this work, we propose cyclic block coordinate methods for nonconvex optimization problems with non-asymptotic gradient norm guarantees. Our convergence analysis is based on a gradient Lipschitz condition with respect to a Mahalanobis norm, inspired by a recent progress on cyclic block coordinate methods. In deterministic settings, our convergence guarantee matches the guarantee of (full-gradient) gradient descent, but with the gradient Lipschitz constant being defined w.r.t. a Mahalanobis norm. In stochastic settings, we use recursive variance reduction to decrease the per-iteration cost and match the arithmetic operation complexity of current optimal stochastic full-gradient methods, with a unified analysis for both finite-sum and infinite-sum cases. We prove a faster linear convergence result when a Polyak-Łojasiewicz (PŁ) condition holds. To our knowledge, this work is the first to provide non-asymptotic convergence guarantees — variance-reduced or not — for a cyclic block coordinate method in general composite (smooth + nonsmooth) nonconvex settings. Our experimental results demonstrate the efficacy of the proposed cyclic scheme in training deep neural nets.

Finding an approximate second-order stationary point (SOSP) is a well-studied and fundamental problem in stochastic nonconvex optimization with many applications in machine learning. However, this problem is poorly understood in the presence of outliers, limiting the use of existing nonconvex algorithms in adversarial settings. In this paper, we study the problem of finding SOSPs in the strong contamination model, where a constant fraction of datapoints are arbitrarily corrupted. We introduce a general framework for efficiently finding an approximate SOSP with \emph{dimension-independent} accuracy guarantees, using $\widetilde{O}({D^2}/{\epsilon})$ samples where $D$ is the ambient dimension and $\epsilon$ is the fraction of corrupted datapoints. As a concrete application of our framework, we apply it to the problem of low rank matrix sensing, developing efficient and provably robust algorithms that can tolerate corruptions in both the sensing matrices and the measurements. In addition, we establish a Statistical Query lower bound providing evidence that the quadratic dependence on $D$ in the sample complexity is necessary for computationally efficient algorithms.

We describe a line-search algorithm which achieves the best-known worst-case complexity results for problems with a certain “strict saddle” property that has been observed to hold in low-rank matrix optimization problems. Our algorithm is adaptive, in the sense that it makes use of backtracking line searches and does not require prior knowledge of the parameters that define the strict saddle property.

We study a class of generalized linear programs (GLP) in a large-scale setting, which includes simple, possibly nonsmooth convex regularizer and simple convex set constraints. By reformulating (GLP) as an equivalent convex-concave min-max problem, we show that the linear structure in the problem can be used to design an efficient, scalable first-order algorithm, to which we give the name Coordinate Linear Variance Reduction (CLVR; pronounced ``clever''). CLVR yields improved complexity results for (GLP) that depend on the max row norm of the linear constraint matrix in (GLP) rather than the spectral norm. When the regularization terms and constraints are separable, CLVR admits an efficient lazy update strategy that makes its complexity bounds scale with the number of nonzero elements of the linear constraint matrix in (GLP) rather than the matrix dimensions. On the other hand, for the special case of linear programs, by exploiting sharpness, we propose a restart scheme for CLVR to obtain empirical linear convergence. Then we show that Distributionally Robust Optimization (DRO) problems with ambiguity sets based on both $f$-divergence and Wasserstein metrics can be reformulated as (GLPs) by introducing sparsely connected auxiliary variables. We complement our theoretical guarantees with numerical experiments that verify our algorithm's practical effectiveness, in terms of wall-clock time and number of data passes.

Finding parameters in a deep neural network (NN) that fit training data is a nonconvex optimization problem, but a basic first-order optimization method (gradient descent) finds a global optimizer with perfect fit (zero-loss) in many practical situations. We examine this phenomenon for the case of Residual Neural Networks (ResNet) with smooth activation functions in a limiting regime in which both the number of layers (depth) and the number of weights in each layer (width) go to infinity. First, we use a mean-field-limit argument to prove that the gradient descent for parameter training becomes a gradient flow for a probability distribution that is characterized by a partial differential equation (PDE) in the large-NN limit. Next, we show that under certain assumptions, the solution to the PDE converges in the training time to a zero-loss solution. Together, these results suggest that the training of the ResNet gives a near-zero loss if the ResNet is large enough. We give estimates of the depth and width needed to reduce the loss below a given threshold, with high probability.

The Underdamped Langevin Monte Carlo (ULMC) is a popular Markov chain Monte Carlo sampling method. It requires the computation of the full gradient of the log-density at each iteration, an expensive operation if the dimension of the problem is high. We propose a sampling method called Random Coordinate ULMC (RC-ULMC), which selects a single coordinate at each iteration to be updated and leaves the other coordinates untouched. We investigate the computational complexity of RC-ULMC and compare it with the classical ULMC for strongly log-concave probability distributions. We show that RC-ULMC is always cheaper than the classical ULMC, with a significant cost reduction when the problem is highly skewed and high dimensional. Our complexity bound for RC-ULMC is also tight in terms of dimension dependence.

Langevin Monte Carlo (LMC) is a popular Markov chain Monte Carlo sampling method. One drawback is that it requires the computation of the full gradient at each iteration, an expensive operation if the dimension of the problem is high. We propose a new sampling method: Random Coordinate LMC (RC-LMC). At each iteration, a single coordinate is randomly selected to be updated by a multiple of the partial derivative along this direction plus noise, while all other coordinates remain untouched. We investigate the total complexity of RC-LMC and compare it with the classical LMC for log-concave probability distributions. We show that when the gradient of the log-density is Lipschitz, RC-LMC is less expensive than the classical LMC if the log-density is highly skewed for high dimensional problems. Further, when both the gradient and the Hessian of the log-density are Lipschitz, RC-LMC is always cheaper than the classical LMC, by a factor proportional to the square root of the problem dimension. In the latter case, we use an example to demonstrate that our estimate of complexity is sharp with respect to the dimension.

Structured nonsmooth convex finite-sum optimization appears in many machine learning applications, including support vector machines and least absolute deviation. For the primal-dual formulation of this problem, we propose a novel algorithm called \emph{Variance Reduction via Primal-Dual Accelerated Dual Averaging (\vrpda)}. In the nonsmooth and general convex setting, \vrpda has the overall complexity $O(nd\log\min \{1/\epsilon, n\} + d/\epsilon )$ in terms of the primal-dual gap, where $n$ denotes the number of samples, $d$ the dimension of the primal variables, and $\epsilon$ the desired accuracy. In the nonsmooth and strongly convex setting, the overall complexity of \vrpda becomes $O(nd\log\min\{1/\epsilon, n\} + d/\sqrt{\epsilon})$ in terms of both the primal-dual gap and the distance between iterate and optimal solution. Both these results for \vrpda improve significantly on state-of-the-art complexity estimates—which are $O(nd\log \min\{1/\epsilon, n\} + \sqrt{n}d/\epsilon)$ for the nonsmooth and general convex setting and $O(nd\log \min\{1/\epsilon, n\} + \sqrt{n}d/\sqrt{\epsilon})$ for the nonsmooth and strongly convex setting—with a simpler and more straightforward algorithm and analysis. Moreover, both complexities are better than \emph{lower} bounds for general convex finite-sum optimization, because our approach makes use of additional, commonly occurring structure. Numerical experiments reveal competitive performance of \vrpda compared to state-of-the-art approaches.

We present a blended conditional gradient approach for minimizing a smooth convex function over a polytope P, combining the Frank{–}Wolfe algorithm (also called conditional gradient) with gradient-based steps, different from away steps and pairwise steps, but still achieving linear convergence for strongly convex functions, along with good practical performance. Our approach retains all favorable properties of conditional gradient algorithms, notably avoidance of projections onto P and maintenance of iterates as sparse convex combinations of a limited number of extreme points of P. The algorithm is lazy, making use of inexpensive inexact solutions of the linear programming subproblem that characterizes the conditional gradient approach. It decreases measures of optimality (primal and dual gaps) rapidly, both in the number of iterations and in wall-clock time, outperforming even the lazy conditional gradient algorithms of Braun et al. 2017. We also present a streamlined version of the algorithm that applies when P is the probability simplex.

We introduce the bilinear bandit problem with low-rank structure in which an action takes the form of a pair of arms from two different entity types, and the reward is a bilinear function of the known feature vectors of the arms. The unknown in the problem is a $d_1$ by $d_2$ matrix $\mathbf{\Theta}^*$ that defines the reward, and has low rank $r \ll \min\{d_1,d_2\}$. Determination of $\mathbf{\Theta}^*$ with this low-rank structure poses a significant challenge in finding the right exploration-exploitation tradeoff. In this work, we propose a new two-stage algorithm called “Explore-Subspace-Then-Refine” (ESTR). The first stage is an explicit subspace exploration, while the second stage is a linear bandit algorithm called “almost-low-dimensional OFUL” (LowOFUL) that exploits and further refines the estimated subspace via a regularization technique. We show that the regret of ESTR is $\widetilde{\mathcal{O}}((d_1+d_2)^{3/2} \sqrt{r T})$ where $\widetilde{\mathcal{O}}$ hides logarithmic factors and $T$ is the time horizon, which improves upon the regret of $\widetilde{\mathcal{O}}(d_1d_2\sqrt{T})$ attained for a naïve linear bandit reduction. We conjecture that the regret bound of ESTR is unimprovable up to polylogarithmic factors, and our preliminary experiment shows that ESTR outperforms a naïve linear bandit reduction.

It is well known that both gradient descent and stochastic coordinate descent achieve a global convergence rate of $O(1/k)$ in the objective value, when applied to a scheme for minimizing a Lipschitz-continuously differentiable, unconstrained convex function. In this work, we improve this rate to $o(1/k)$. We extend the result to proximal gradient and proximal coordinate descent on regularized problems to show similar $o(1/k)$ convergence rates. The result is tight in the sense that a rate of $O(1/k^{1+\epsilon})$ is not generally attainable for any $\epsilon>0$, for any of these methods.

Prediction of compounds that are active against a desired biological target is a common step in drug discovery efforts. Virtual screening methods seek some active-enriched fraction of a library for experimental testing. Where data are too scarce to train supervised learning models for compound prioritization, initial screening must provide the necessary data. Commonly, such an initial library is selected on the basis of chemical diversity by some pseudo-random process (for example, the first few plates of a larger library) or by selecting an entire smaller library. These approaches may not produce a sufficient number or diversity of actives. An alternative approach is to select an informer set of screening compounds on the basis of chemogenomic information from previous testing of compounds against a large number of targets. We compare different ways of using chemogenomic data to choose a small informer set of compounds based on previously measured bioactivity data. We develop this Informer-Based-Ranking (IBR) approach using the Published Kinase Inhibitor Sets (PKIS) as the chemogenomic data to select the informer sets. We test the informer compounds on a target that is not part of the chemogenomic data, then predict the activity of the remaining compounds based on the experimental informer data and the chemogenomic data. Through new chemical screening experiments, we demonstrate the utility of IBR strategies in a prospective test on three kinase targets not included in the PKIS.

Training set bugs are flaws in the data that adversely affect machine learning. The training set is usually too large for manual inspection, but one may have the resources to verify a few trusted items. The set of trusted items may not by itself be adequate for learning, so we propose an algorithm that uses these items to identify bugs in the training set and thus improves learning. Specifically, our approach seeks the smallest set of changes to the training set labels such that the model learned from this corrected training set predicts labels of the trusted items correctly. We flag the items whose labels are changed as potential bugs, whose labels can be checked for veracity by human experts. To find the bugs in this way is a challenging combinatorial bilevel optimization problem, but it can be relaxed into a continuous optimization problem.Experiments on toy and real data demonstrate that our approach can identify training set bugs effectively and suggest appropriate changes to the labels. Our algorithm is a step toward trustworthy machine learning.

Techniques for reducing the variance of gradient estimates used in stochastic programming algorithms for convex finite-sum problems have received a great deal of attention in recent years. By leveraging dissipativity theory from control, we provide a new perspective on two important variance-reduction algorithms: SVRG and its direct accelerated variant Katyusha. Our perspective provides a physically intuitive understanding of the behavior of SVRG-like methods via a principle of energy conservation. The tools discussed here allow us to automate the convergence analysis of SVRG-like methods by capturing their essential properties in small semidefinite programs amenable to standard analysis and computational techniques. Our approach recovers existing convergence results for SVRG and Katyusha and generalizes the theory to alternative parameter choices. We also discuss how our approach complements the linear coupling technique. Our combination of perspectives leads to a better understanding of accelerated variance-reduced stochastic methods for finite-sum problems.

We propose a communication- and computation-efficient distributed optimization algorithm using second-order information for solving ERM problems with a nonsmooth regularization term. Current second-order and quasi-Newton methods for this problem either do not work well in the distributed setting or work only for specific regularizers. Our algorithm uses successive quadratic approximations, and we describe how to maintain an approximation of the Hessian and solve subproblems efficiently in a distributed manner. The proposed method enjoys global linear convergence for a broad range of non-strongly convex problems that includes the most commonly used ERMs, thus requiring lower communication complexity. It also converges on non-convex problems, so has the potential to be used on applications such as deep learning. Initial computational results on convex problems demonstrate that our method significantly improves on communication cost and running time over the current state-of-the-art methods.

Distributed model training suffers from communication overheads due to frequent gradient updates transmitted between compute nodes. To mitigate these overheads, several studies propose the use of sparsified stochastic gradients. We argue that these are facets of a general sparsification method that can operate on any possible atomic decomposition. Notable examples include element-wise, singular value, and Fourier decompositions. We present ATOMO, a general framework for atomic sparsification of stochastic gradients. Given a gradient, an atomic decomposition, and a sparsity budget, ATOMO gives a random unbiased sparsification of the atoms minimizing variance. We show that recent methods such as QSGD and TernGrad are special cases of ATOMO, and that sparsifiying the singular value decomposition of neural networks gradients, rather than their coordinates, can lead to significantly faster distributed training.

This paper describes a new parameter-free online learning algorithm for changing environments. In comparing against algorithms with the same time complexity as ours, we obtain a strongly adaptive regret bound that is a factor of at least $\sqrt\log(T)$ better, where $T$ is the time horizon. Empirical results show that our algorithm outperforms state-of-the-art methods in learning with expert advice and metric learning scenarios.

The k-support and OWL norms generalize the l1 norm, providing better prediction accuracy and better handling of correlated variables. We study the norms obtained from extending the k-support norm and OWL norms to the setting in which there are overlapping groups. The resulting norms are in general NP-hard to compute, but they are tractable for certain collections of groups. To demonstrate this fact, we develop a dynamic program for the problem of projecting onto the set of vectors supported by a fixed number of groups. Our dynamic program utilizes tree decompositions and its complexity scales with the treewidth. This program can be converted to an extended formulation which, for the associated group structure, models the k-group support norms and an overlapping group variant of the ordered weighted l1 norm. Numerical results demonstrate the efficacy of the new penalties.

We introduce a simple, efficient method that improves stochastic policies for Markov decision processes. The computational complexity is the same as that of the value estimation problem. We prove that when the value estimation error is small, this method gives an improvement in performance that increases with certain variance properties of the initial policy and transition dynamics. Performance in numerical experiments compares favorably with previous policy improvement algorithms.

The problem of projecting onto the permutahedron \mathbfPH(c) – the convex hull of all permutations of a fixed vector c – under a uniformly separable Bregman divergence is shown to be reducible to the Isotonic Optimization problem. This allows us to employ known fast algorithms to improve on several recent results on Bregman projections onto permutahedra. In addition, we present a new algorithm \rm mergepool that have better complexity when the number of distinct entries d in the vector c is small, the simplex being one such example, with c=(1,0,0,\dotsc,0)^T and d=2. \rm mergepool runs in O(n \log d) for certain popular Bregman divergence measures and requires O((n \log d) \log \fracU ε) to find ε-close solutions for general uniformly separable Bregman divergences, where U is a bound on the width of the interval containing the dual solution components. These estimates matches or improves best known bounds for all Bregman projection problems onto various permutahedra, including recent results for projection onto the simplex. The same complexity bounds apply to signed permutahedra, a class that includes the \ell_1-ball as a special case. In summary, this work describes a fast unified approach to this well-known class of problems.

This special issue highlights a number of algorithmic approaches that are fundamental to data analysis, both in formulating and solving problems that relate to Big Data.

We describe an asynchronous parallel stochastic coordinate descent algorithm for minimizing smooth unconstrained or separably constrained functions. The method achieves a linear convergence rate on functions that satisfy an essential strong convexity property and a sublinear rate (1/K) on general convex functions. Near-linear speedup on a multicore system can be expected if the number of processors is O(n^1/2) in unconstrained optimization and O(n^1/4) in the separable-constrained case, where n is the number of variables. We describe results from implementation on 40-core processors.

We present a family of online algorithms for real-time factorization-based structure from motion, leveraging a relationship between the incremental singular value decomposition and recent work in online matrix completion. Our methods are orders of magnitude faster than previous state of the art, can handle missing data and a variable number of feature points, and are robust to noise and sparse outliers. Experiments show that they perform well in both online and batch settings. We also provide an implementation which is able to produce 3D models in real time using a laptop with a webcam.

In many applications in signal and image processing, communications, and system identification, one aims to recover a signal that has a simple representation in a given basis or frame. Key devices for obtaining such representations are objects called atoms, and functions called atomic norms. These concepts unify the idea of simple representations across several known applications, and motivate extensions to new problem classes of interest. In important special cases, fast and efficient algorithms are available to solve the reconstruction problems, but an approach that works well for the general atomic-norm paradigm has not been forthcoming to date. In this paper, we combine a greedy selection scheme with a backward step that sparsifies the basis by removing less significant elements that were included at earlier iterations. We show that the overall scheme achieves the same convergence rate as the forward greedy scheme alone, provided that backward steps are taken only when they do not degrade the solution quality too badly. Finally, we validate our method by describing applications to three problems of interest.

Optimization tools are vital to data analysis and learning. The optimization perspective has provided valuable insights, and optimization formulations have led to practical algorithms with good theoretical properties. In turn, the rich collection of problems in learning and data analysis is providing fresh perspectives on optimization algorithms and is driving new fundamental research in the area. We discuss research on several areas in this domain, including signal reconstruction, manifold learning, and regression/classification, describing in each case recent research in which optimization algorithms have been developed and applied successfully. A particular focus is asynchronous parallel algorithms for optimization and linear algebra, and their applications in data analysis and learning.

Many problems in machine learning can be solved by rounding the solution of an appropriate linear program. We propose a scheme that is based on a quadratic program relaxation which allows us to use parallel stochastic-coordinate-descent to approximately solve large linear programs efficiently. Our software is an order of magnitude faster than Cplex (a commercial linear programming solver) and yields similar solution quality. Our results include a novel perturbation analysis of a quadratic-penalty formulation of linear programming and a convergence result, which we use to derive running time and quality guarantees.

Optimization techniques have been used for many years in the formulation and solution of computational problems arising in speech and language processing. Such techniques are found in the Baum-Welch, extended Baum-Welch (EBW), Rprop, and GIS algorithms, for example. Additionally, the use of regularization terms has been seen in other applications of sparse optimization. This paper outlines a range of problems in which optimization formulations and algorithms play a role, giving some additional details on certain application problems in machine translation, speaker/language recognition, and automatic speech recognition. Several approaches developed in the speech and language processing communities are described in a way that makes them more recognizable as optimization procedures. Our survey is not exhaustive and is complemented by other papers in this volume.

Over the past few decades, a variety of specialized approaches have been proposed to solve large problems in speech recognition. Conventional optimization techniques have not been widely applied, because the problems do not readily admit an objective for evaluating a given set of parameters and because of the large number of parameters. This situation is changing, due to recent developments in algorithmic optimization. In this paper, we review the specialized algorithms, including methods derived from the extended Baum-Welch (EBW) approach, Rprop, and GIS. We discuss optimization frameworks that could also potentially be applied, and outline some connections between the optimization methods and existing specialized methods.

BACKGROUND: In systems biology, the task of reverse engineering gene pathways from data has been limited not just by the curse of dimensionality (the interaction space is huge) but also by systematic error in the data. The gene expression barcode reduces spurious association driven by batch effects and probe effects. The binary nature of the resulting expression calls lends itself perfectly to modern regularization approaches that thrive in high-dimensional settings. RESULTS: The Partitioned LASSO-Patternsearch algorithm is proposed to identify patterns of multiple dichotomous risk factors for outcomes of interest in genomic studies. A partitioning scheme is used to identify promising patterns by solving many LASSO-Patternsearch subproblems in parallel. All variables that survive this stage proceed to an aggregation stage where the most significant patterns are identified by solving a reduced LASSO-Patternsearch problem in just these variables. This approach was applied to genetic data sets with expression levels dichotomized by gene expression bar code. Most of the genes and second-order interactions thus selected and are known to be related to the outcomes. CONCLUSIONS: We demonstrate with simulations and data analyses that the proposed method not only selects variables and patterns more accurately, but also provides smaller models with better prediction accuracy, in comparison to several alternative methodologies.

OBJECTIVE: To describe an analytical framework for quantifying the societal savings and financial consequences of a health information exchange (HIE), and to demonstrate its use in designing pricing policies for sustainable HIEs. MATERIALS AND METHODS: We developed a linear programming model to (1) quantify the financial worth of HIE information to each of its participating institutions and (2) evaluate three HIE pricing policies: fixed-rate annual, charge per visit, and charge per look-up. We considered three desired outcomes of HIE-related emergency care (modeled as parameters): preventing unrequired hospitalizations, reducing duplicate tests, and avoiding emergency department (ED) visits. We applied this framework to 4639 ED encounters over a 12-month period in three large EDs in Milwaukee, Wisconsin, using Medicare/Medicaid claims data, public reports of hospital admissions, published payer mix data, and use data from a not-for-profit regional HIE. RESULTS: For this HIE, data accesses produced net financial gains for all providers and payers. Gains, due to HIE, were more significant for providers with more health maintenance organizations patients. Reducing unrequired hospitalizations and avoiding repeat ED visits were responsible for more than 70% of the savings. The results showed that fixed annual subscriptions can sustain this HIE, while ensuring financial gains to all participants. Sensitivity analysis revealed that the results were robust to uncertainties in modeling parameters. DISCUSSION: Our specific HIE pricing recommendations depend on the unique characteristics of this study population. However, our main contribution is the modeling approach, which is broadly applicable to other populations.

Statistical dependencies among wavelet coefficients are commonly represented by graphical models such as hidden Markov trees (HMTs). However, in linear inverse problems such as deconvolution, tomography, and compressed sensing, the presence of a sensing or observation matrix produces a linear mixing of the simple Markovian dependency structure. This leads to reconstruction problems that are non-convex optimizations. Past work has dealt with this issue by resorting to greedy or suboptimal iterative reconstruction methods. In this paper, we propose new modeling approaches based on group-sparsity penalties that leads to convex optimizations that can be solved exactly and efficiently. We show that the methods we develop perform significantly better in de-convolution and compressed sensing applications, while being as computationally efficient as standard coefficient-wise approaches such as lasso.

Stochastic Gradient Descent (SGD) is a popular algorithm that can achieve state-of-the-art performance on a variety of machine learning tasks. Several researchers have recently proposed schemes to parallelize SGD, but all require performance-destroying memory locking and synchronization. This work aims to show using novel theoretical analysis, algorithms, and implementation that SGD can be implemented without any locking. We present an update scheme called Hogwild which allows processors access to shared memory with the possibility of overwriting each other's work. We show that when the associated optimization problem is sparse, meaning most gradient updates only modify small parts of the decision variable, then Hogwild achieves a nearly optimal rate of convergence. We demonstrate experimentally that Hogwild outperforms alternative schemes that use locking by an order of magnitude.

The goal of the sparse approximation problem is to approximate a target signal using a linear combination of a few elementary signals drawn from a fixed collection. This paper surveys the major practical algorithms for sparse approximation. Specific attention is paid to computational issues, to the circumstances in which individual methods tend to perform well, and to the theoretical guarantees available. Many fundamental questions in electrical engineering, statistics, and applied mathematics can be posed as sparse approximation problems, making these algorithms versatile and relevant to a plethora of applications.

Finding sparse approximate solutions to large underdetermined linear systems of equations is a common problem in signal/image processing and statistics. Basis pursuit, the least absolute shrinkage and selection operator (LASSO), wavelet-based deconvolution and reconstruction, and compressed sensing (CS) are a few well-known areas in which problems of this type appear. One standard approach is to minimize an objective function that includes a quadratic (pound2) error term added to a sparsity-inducing (usuallypound1) regularizer. We present an algorithmic framework for the more general problem of minimizing the sum of a smooth convex function and a nonsmooth, possibly nonconvex, sparsity-inducing function. We propose iterative methods in which each step is an optimization subproblem involving a separable quadratic term (diagonal Hessian) plus the original sparsity-inducing term. Our approach is suitable for cases in which this subproblem can be solved much more rapidly than the original problem. In addition to solving the standardpound2-pound1case, our approach handles other problems, e.g.,poundpregularizers with p ne 1, or group-separable (GS) regularizers. Experiments with CS problems show that our approach provides state-of-the-art speed for the standardpound2-pound1problem, and is also efficient on problems with GS regularizers.

Exponential bandwidth scaling has been a fundamental driver of the growth and popularity of the Internet. However, increases in bandwidth have been accompanied by increases in power consumption, and despite sustained system design efforts to address power demand, significant technological challenges remain that threaten to slow future bandwidth growth. In this paper we describe the power and associated heat management challenges in today's routers. We advocate a broad approach to addressing this problem that includes making power-awareness a primary objective in the design and configuration of networks, and in the design and implementation of network protocols. We support our arguments by providing a case study of power demands of two standard router platforms that enables us to create a generic model for router power consumption. We apply this model in a set of target network configurations and use mixed integer optimization techniques to investigate power consumption, performance and robustness in static network design and in dynamic routing. Our results indicate the potential for significant power savings in operational networks by including power-awareness.

MOTIVATION: For many biotechnological purposes, it is desirable to redesign proteins to be more structurally and functionally stable at higher temperatures. For example, chemical reactions are intrinsically faster at higher temperatures, so using enzymes that are stable at higher temperatures would lead to more efficient industrial processes. We describe an innovative and computationally efficient method called Improved Configurational Entropy (ICE), which can be used to redesign a protein to be more thermally stable (i.e. stable at high temperatures). This can be accomplished by systematically modifying the amino acid sequence via local structural entropy (LSE) minimization. The minimization problem is modeled as a shortest path problem in an acyclic graph with nonnegative weights and is solved efficiently using Dijkstra's method.

An optimization framework for three-dimensional conformal radiation therapy is presented. In conformal therapy, beams of radiation are applied to a patient from different directions, where the aperture through which the beam is delivered from each direction is chosen to match the shape of the tumor, as viewed from that direction. Wedge filters may be used to produce a gradient in beam intensity across the aperture. Given a set of equispaced beam angles, a mixed-integer linear program can be solved to determine the most effective angles to be used in a treatment plan, the weight (exposure time) to be used for each beam, and the type and orientation of wedges to be used. Practical solution techniques for this problem are described; they include strengthening of the formulation and solution of smaller approximate problems obtained by a reduced parametrization of the treatment region. In addition, techniques for controlling the dose-volume histogram implicitly for various parts of the treatment region using hot- and cold-spot control parameters are presented. Computational results are given that show the effectiveness of the proposed approach on practical data sets.

Data streaming systems face the possibility of having to shed load in the case of CPU or memory resource limitations. We study the CPU limited scenario in detail. First, we propose a new model for the CPU cost. Then we formally state the problem of shedding load for the goal of obtaining the maximum possible subset of the complete answer, and propose an online strategy for semantic load shedding. Moving on to random load shedding, we discuss random load shedding strategies that decouple the window maintenance and tuple production operations of the symmetric hash join, and prove that one of them — Probe-No-Insert — always dominates the previously proposed coin flipping strategy.

We introduce the problem of labeling a particle's mass spectrum with the substances it contains, and develop several formal representations of the problem, taking into account practical complications such as unknown compounds and noise. This task is currently a bottle-neck in analyzing data from a new generation of instruments for real-time environmental monitoring.

Recent scalable multicast streaming protocols for on-demand delivery of media content offer the promise of greatly reduced server and network bandwidth. However, a key unresolved issue is how to design scalable content distribution systems that place replica servers closer to various client populations and route client requests and response streams so as to minimize the total server and network delivery cost. This issue is significantly more complex than the design of distribution systems for traditional Web files or unicast on-demand streaming, for two reasons. First, closest server and shortest path routing does not minimize network bandwidth usage; instead, the optimal routing of client requests and server multicasts is complex and interdependent. Second, the server bandwidth usage increases with the number of replicas. Nevertheless, this paper shows that the complex replica placement and routing optimization problem, in its essential form, can be expressed fairly simply, and can be solved for example client populations and realistic network topologies. The solutions show that the optimal scalable system can differ significantly from the optimal system for conventional delivery. Furthermore, simple canonical networks are analyzed to develop insights into effective heuristics for near-optimal placement and routing. The proposed new heuristics can be used for designing large and heterogeneous systems that are of practical interest. For a number of example networks, the best heuristics produce systems with total delivery cost that is within 16% of optimality.

The object-oriented software package OOQP for solving convex quadratic programming problems (QP) is described. The primal-dual interior point algorithms supplied by OOQP are implemented in a way that is largely independent of the problem structure. Users may exploit problem structure by supplying linear algebra, problem data, and variable classes that are customized to their particular applications. The OOQP distribution contains default implementations that solve several important QP problem types, including general sparse and dense QPs, bound-constrained QPs, and QPs arising from support vector machines and Huber regression. The implementations supplied with the OOQP distribution are based on such well known linear algebra packages as MA27/57, LAPACK, and PETSc. OOQP demonstrates the usefulness of object-oriented design in optimization software development, and establishes standards that can be followed in the design of software packages for other classes of optimization problems. A number of the classes in OOQP may also be reusable directly in other codes.

This paper develops a performance model that is used to control the adaptive execution the ATR code for solving large stochastic optimization problems on computational grids. A detailed analysis of the execution characteristics of ATR is used to construct the performance model that is then used to specify (a) near-optimal dynamic values of parameters that govern the distribution of work, and (b) a new task scheduling algorithm. Together, these new features minimize ATR execution time on any collection of compute nodes, including a varying collection of heterogeneous nodes. The new adaptive code runs up to eight-fold faster than the previously optimized code, and requires no input parameters from the user to guide the distribution of work. Furthermore, the modeling process led to several changes in the Condor runtime environment, including the new task scheduling algorithm, that produce significant performance improvements for master-worker computations as well as possibly other types of grid applications.