Motivated by distributed machine learning settings such as Federated Learning, we consider the problem of fitting a statistical model across a distributed collection of heterogeneous data sets whose similarity structure is encoded by a graph topology. Precisely, we analyse the case where each node is associated with fitting a sparse linear model, and edges join two nodes if the difference of their solutions is also sparse. We propose a method based on Basis Pursuit Denoising with a total variation penalty, and provide finite sample guarantees for sub-Gaussian design matrices. Taking the root of the tree as a reference node, we show that if the sparsity of the differences across nodes is smaller than the sparsity at the root, then recovery is successful with fewer samples than by solving the problems independently, or by using methods that rely on a large overlap in the signal supports, such as the group Lasso. We consider both the noiseless and noisy setting, and numerically investigate the performance of distributed methods based on Distributed Alternating Direction Methods of Multipliers (ADMM) and hyperspectral unmixing.

We study estimation of a gradient-sparse parameter vector $\boldsymbol{\theta}^* \in \mathbb{R}^p$, having strong gradient-sparsity $s^*:=\|\nabla_G \boldsymbol{\theta}^*\|_0$ on an underlying graph $G$. Given observations $Z_1,\ldots,Z_n$ and a smooth, convex loss function $\mathcal{L}$ for which $\boldsymbol{\theta}^*$ minimizes the population risk $\mathbb{E}[\mathcal{L}(\boldsymbol{\theta};Z_1,\ldots,Z_n)]$, we propose to estimate $\boldsymbol{\theta}^*$ by a projected gradient descent algorithm that iteratively and approximately projects gradient steps onto spaces of vectors having small gradient-sparsity over low-degree spanning trees of $G$. We show that, under suitable restricted strong convexity and smoothness assumptions for the loss, the resulting estimator achieves the squared-error risk $\frac{s^*}{n} \log (1+\frac{p}{s^*})$ up to a multiplicative constant that is independent of $G$. In contrast, previous polynomial-time algorithms have only been shown to achieve this guarantee in more specialized settings, or under additional assumptions for $G$ and/or the sparsity pattern of $\nabla_G \boldsymbol{\theta}^*$. As applications of our general framework, we apply our results to the examples of linear models and generalized linear models with random design.

Feature selection problems have been extensively studied in the setting of linear estimation (e.g. LASSO), but less emphasis has been placed on feature selection for non-linear functions. In this study, we propose a method for feature selection in neural network estimation problems. The new procedure is based on probabilistic relaxation of the $\ell_0$ norm of features, or the count of the number of selected features. Our $\ell_0$-based regularization relies on a continuous relaxation of the Bernoulli distribution; such relaxation allows our model to learn the parameters of the approximate Bernoulli distributions via gradient descent. The proposed framework simultaneously learns either a nonlinear regression or classification function while selecting a small subset of features. We provide an information-theoretic justification for incorporating Bernoulli distribution into feature selection. Furthermore, we evaluate our method using synthetic and real-life data to demonstrate that our approach outperforms other commonly used methods in both predictive performance and feature selection.

We investigate the feasibility of learning from both fully-labeled supervised data and contextual bandit data. We specifically consider settings in which the underlying learning signal may be different between these two data sources. Theoretically, we state and prove no-regret algorithms for learning that is robust to divergences between the two sources. Empirically, we evaluate some of these algorithms on a large selection of datasets, showing that our approaches are feasible, and helpful in practice.

When tracking user-specific online activities, each user's preference is revealed in the form of choices and comparisons. For example, a user's purchase history is a record of her choices, i.e. which item was chosen among a subset of offerings. A user's preferences can be observed either explicitly as in movie ratings or implicitly as in viewing times of news articles. Given such individualized ordinal data in the form of comparisons and choices, we address the problem of collaboratively learning representations of the users and the items. The learned features can be used to predict a user's preference of an unseen item to be used in recommendation systems. This also allows one to compute similarities among users and items to be used for categorization and search. Motivated by the empirical successes of the MultiNomial Logit (MNL) model in marketing and transportation, and also more recent successes in word embedding and crowdsourced image embedding, we pose this problem as learning the MNL model parameters that best explain the data. We propose a convex relaxation for learning the MNL model, and show that it is minimax optimal up to a logarithmic factor by comparing its performance to a fundamental lower bound. This characterizes the minimax sample complexity of the problem, and proves that the proposed estimator cannot be improved upon other than by a logarithmic factor. Further, the analysis identifies how the accuracy depends on the topology of sampling via the spectrum of the sampling graph. This provides a guideline for designing surveys when one can choose which items are to be compared. This is accompanied by numerical simulations on synthetic and real data sets, confirming our theoretical predictions.

Greedy algorithms are widely used for problems in machine learning such as feature selection and set function optimization. Unfortunately, for large datasets, the running time of even greedy algorithms can be quite high. This is because for each greedy step we need to refit a model or calculate a function using the previously selected choices and the new candidate. Two algorithms that are faster approximations to the greedy forward selection were introduced recently [Mirzasoleiman et al., 2013, 2015]. They achieve better performance by exploiting stochastic evaluation and distributed computation respectively. Both algorithms have provable performance guarantees for submodular functions. In this paper we show that divergent from previously held opinion, submodularity is not required to obtain approximation guarantees for these two algorithms. Specifically, we show that a generalized concept of weak submodularity suffices to give multiplicative approximation guarantees. Our result extends the applicability of these algorithms to a larger class of functions. Furthermore, we show that a bounded submodularity ratio can be used to provide data dependent bounds that can sometimes be tighter also for submodular functions. We empirically validate our work by showing superior performance of fast greedy approximations versus several established baselines on artificial and real datasets.

We provide new approximation guarantees for greedy low rank matrix estimation under standard assumptions of restricted strong convexity and smoothness. Our novel analysis also uncovers previously unknown connections between the low rank estimation and combinatorial optimization, so much so that our bounds are reminiscent of corresponding approximation bounds in submodular maximization. Additionally, we provide also provide statistical recovery guarantees. Finally, we present empirical comparison of greedy estimation with established baselines on two important real-world problems.

The inverse covariance matrix provides considerable insight for understanding statistical models in the multivariate setting. In particular, when the distribution over variables is assumed to be multivariate normal, the sparsity pattern in the inverse covariance matrix, commonly referred to as the precision matrix, corresponds to the adjacency matrix representation of the Gauss-Markov graph, which encodes conditional independence statements between variables. Minimax results under the spectral norm have previously been established for covariance matrices, both sparse and banded, and for sparse precision matrices. We establish minimax estimation bounds for estimating banded precision matrices under the spectral norm. Our results greatly improve upon the existing bounds; in particular, we find that the minimax rate for estimating banded precision matrices matches that of estimating banded covariance matrices. The key insight in our analysis is that we are able to obtain barely-noisy estimates of $k \times k$ subblocks of the precision matrix by inverting slightly wider blocks of the empirical covariance matrix along the diagonal. Our theoretical results are complemented by experiments demonstrating the sharpness of our bounds.

Electronic health records (EHR) provide opportunities to leverage vast arrays of data to help prevent adverse events, improve patient outcomes, and reduce hospital costs. This paper develops a postoperative complications prediction system by extracting data from the EHR and creating features. The analytic engine then provides model accuracy, calibration, feature ranking, and personalized feature responses. This allows clinicians to interpret the likelihood of an adverse event occurring, general causes for these events, and the contributing factors for each specific patient. The patient cohort considered was 5214 patients in Yale-New Haven Hospital undergoing major cardiovascular procedures. Cohort-specific models predicted the likelihood of postoperative respiratory failure and infection, and achieved an area under the receiver operating characteristic curve of 0.81 for respiratory failure and 0.83 for infection.

We consider a serious, previously-unexplored challenge facing almost all approaches to scaling up entity resolution (ER) to multiple data sources: the prohibitive cost of labeling training data for supervised learning of similarity scores for each pair of sources. While there exists a rich literature describing almost all aspects of pairwise ER, this new challenge is arising now due to the unprecedented ability to acquire and store data from online sources, interest in features driven by ER such as enriched search verticals, and the uniqueness of noisy and missing data characteristics for each source. We show on real-world and synthetic data that for state-of-the-art techniques, the reality of heterogeneous sources means that the number of labeled training data must scale quadratically in the number of sources, just to maintain constant precision/recall. We address this challenge with a brand new transfer learning algorithm which requires far less training data (or equivalently, achieves superior accuracy with the same data) and is trained using fast convex optimization. The intuition behind our approach is to adaptively share structure learned about one scoring problem with all other scoring problems sharing a data source in common. We demonstrate that our theoretically-motivated approach improves upon existing techniques for multi-source ER.

We develop and analyze stochastic optimization algorithms for problems in which the expected loss is strongly convex, and the optimum is (approximately) sparse. Previous approaches are able to exploit only one of these two structures, yielding a $\order(\pdim/T)$ convergence rate for strongly convex objectives in $\pdim$ dimensions and $\order(\sqrt{\spindex( \log\pdim)/T})$ convergence rate when the optimum is $\spindex$-sparse. Our algorithm is based on successively solving a series of $\ell_1$-regularized optimization problems using Nesterov's dual averaging algorithm. We establish that the error of our solution after $T$ iterations is at most $\order(\spindex(\log\pdim)/T)$, with natural extensions to approximate sparsity. Our results apply to locally Lipschitz losses including the logistic, exponential, hinge and least-squares losses. By recourse to statistical minimax results, we show that our convergence rates are optimal up to constants. The effectiveness of our approach is also confirmed in numerical simulations where we compare to several baselines on a least-squares regression problem.

The question of aggregating pairwise comparisons to obtain a global ranking over a collection of objects has been of interest for a very long time: be it ranking of online gamers (e.g. MSR’s TrueSkill system) and chess players, aggregating social opinions, or deciding which product to sell based on transactions. In most settings, in addition to obtaining ranking, finding ‘scores’ for each object (e.g. player’s rating) is of interest to understanding the intensity of the preferences. In this paper, we propose a novel iterative rank aggregation algorithm for discovering scores for objects from pairwise comparisons. The algorithm has a natural random walk interpretation over the graph of objects with edges present between two objects if they are compared; the scores turn out to be the stationary probability of this random walk. The algorithm is model independent. To establish the efficacy of our method, however, we consider the popular Bradley-Terry-Luce (BTL) model in which each object has an associated score which determines the probabilistic outcomes of pairwise comparisons between objects. We bound the finite sample error rates between the scores assumed by the BTL model and those estimated by our algorithm. This, in essence, leads to order-optimal dependence on the number of samples required to learn the scores well by our algorithm. Indeed, the experimental evaluation shows that our (model independent) algorithm performs as well as the Maximum Likelihood Estimator of the BTL model and outperforms a recently proposed algorithm by Ammar and Shah [1].

Given a collection ofr≥ 2 linear regression problems inpdimensions, suppose that the regression coefficients share partially common supports of size at mosts. This set-up suggests the use of ℓ1/ℓ∞-regularized regression for joint estimation of thep×rmatrix of regression coefficients. We analyze the high-dimensional scaling of ℓ1/ℓ∞-regularized quadratic programming, considering both consistency rates in ℓ∞-norm, and how the minimal sample sizenrequired for consistent variable selection scales with model dimension, sparsity, and overlap between the supports. We first establish bounds on the ℓ∞-error as well sufficient conditions for exact variable selection for fixed design matrices, as well as for designs drawn randomly from general Gaussian distributions. Specializing to the caser= 2 linear regression problems with standard Gaussian designs whose supports overlap in a fraction α ∈ [0,1] of their entries, we prove that ℓ1/ℓ∞-regularized method undergoes a phase transition characterized by the rescaled sample size θ1,∞(n,p,s, α) =n/{(4 - 3 α)slog(p-(2- α)s)}. An implication is that the use of ℓ1/ℓ∞-regularization yields improved statistical efficiency if the overlap parameter is large enough ( α >; 2/3), but has worse statistical efficiency than a naive Lasso-based approach for moderate to small overlap (αℓ1/ℓ∞block regularization: if the data does not match its structure very closely, it can impair statistical performance relative to computationally less expensive schemes.

Many statistical $M$-estimators are based on convex optimization problems formed by the weighted sum of a loss function with a norm-based regularizer. We analyze the convergence rates of first-order gradient methods for solving such problems within a high-dimensional framework that allows the data dimension $d$ to grow with (and possibly exceed) the sample size $n$. This high-dimensional structure precludes the usual global assumptions---namely, strong convexity and smoothness conditions---that underlie classical optimization analysis. We define appropriately restricted versions of these conditions, and show that they are satisfied with high probability for various statistical models. Under these conditions, our theory guarantees that Nesterov's first-order method~\cite{Nesterov07} has a globally geometric rate of convergence up to the statistical precision of the model, meaning the typical Euclidean distance between the true unknown parameter $\theta^*$ and the optimal solution $\widehat{\theta}$. This globally linear rate is substantially faster than previous analyses of global convergence for specific methods that yielded only sublinear rates. Our analysis applies to a wide range of $M$-estimators and statistical models, including sparse linear regression using Lasso ($\ell_1$-regularized regression), group Lasso, block sparsity, and low-rank matrix recovery using nuclear norm regularization. Overall, this result reveals an interesting connection between statistical precision and computational efficiency in high-dimensional estimation.

The estimation of high-dimensional parametric models requires imposing some structure on the models, for instance that they be sparse, or that matrix structured parameters have low rank. A general approach for such structured parametric model estimation is to use regularized M-estimation procedures, which regularize a loss function that measures goodness of fit of the parameters to the data with some regularization function that encourages the assumed structure. In this paper, we aim to provide a unified analysis of such regularized M-estimation procedures. In particular, we report the convergence rates of such estimators in any metric norm. Using just our main theorem, we are able to rederive some of the many existing results, but also obtain a wide range of novel convergence rates results. Our analysis also identifies key properties of loss and regularization functions such as restricted strong convexity, and decomposability, that ensure the corresponding regularized M-estimators have good convergence rates.

We consider the following instance of transfer learning: given a pair of regression problems, suppose that the regression coefficients share a partially common support, parameterized by the overlap fraction $\overlap$ between the two supports. This set-up suggests the use of $1, \infty$-regularized linear regression for recovering the support sets of both regression vectors. Our main contribution is to provide a sharp characterization of the sample complexity of this $1,\infty$ relaxation, exactly pinning down the minimal sample size $n$ required for joint support recovery as a function of the model dimension $\pdim$, support size $\spindex$ and overlap $\overlap \in [0,1]$. For measurement matrices drawn from standard Gaussian ensembles, we prove that the joint $1,\infty$-regularized method undergoes a phase transition characterized by order parameter $\orpar(\numobs, \pdim, \spindex, \overlap) = \numobs{(4 - 3 \overlap) s \log(p-(2-\overlap)s)}$. More precisely, the probability of successfully recovering both supports converges to $1$ for scalings such that $\orpar > 1$, and converges to $0$ to scalings for which $\orpar < 1$. An implication of this threshold is that use of $1, \infty$-regularization leads to gains in sample complexity if the overlap parameter is large enough ($\overlap > 2/3$), but performs worse than a naive approach if $\overlap < 2/3$. We illustrate the close agreement between these theoretical predictions, and the actual behavior in simulations. Thus, our results illustrate both the benefits and dangers associated with block-$1,\infty$ regularization in high-dimensional inference.