Gaussian processes are a natural extension of the ubiquitous kernel regression to the Bayesian setting where the regression parameters are modelled as random variables with a Gaussian prior distribution [1]. Given the training observations, Gaussian processes generate posterior probabilities of the target or output for new inputs or observations, as a function of the training data and the input kernel function [2]. Gaussian process models and its variants have been applied in a number of diverse fields such as model predictive control and system analysis [3]-[7], latent variable models [8]-[11], multi-task learning [10], [12], [13], image analysis and synthesis [14]- [17], speech processing [18]-[20], and magnetic resonance imaging (MRI) [21], [22]. Gaussian processes have also been extended to a non-stationary regression setting [23]-[25] and for regression over complex-valued data [26]. Recently, Gaussian processes were shown to be useful in training and analysis of deep neural networks, and that a Gaussian process can be viewed as a neural network with a single infinite-dimensional layer of hidden units [27], [28].

This paper considers the problem of solving systems of quadratic equations, namely, recovering an object of interest $\mathbf{x}^{\natural}\in\mathbb{R}^{n}$ from $m$ quadratic equations/samples $y_{i}=(\mathbf{a}_{i}^{\top}\mathbf{x}^{\natural})^{2}$, $1\leq i\leq m$. This problem, also dubbed as phase retrieval, spans multiple domains including physical sciences and machine learning. We investigate the efficiency of gradient descent (or Wirtinger flow) designed for the nonconvex least squares problem. We prove that under Gaussian designs, gradient descent --- when randomly initialized --- yields an $\epsilon$-accurate solution in $O\big(\log n+\log(1/\epsilon)\big)$ iterations given nearly minimal samples, thus achieving near-optimal computational and sample complexities at once. This provides the first global convergence guarantee concerning vanilla gradient descent for phase retrieval, without the need of (i) carefully-designed initialization, (ii) sample splitting, or (iii) sophisticated saddle-point escaping schemes. All of these are achieved by exploiting the statistical models in analyzing optimization algorithms, via a leave-one-out approach that enables the decoupling of certain statistical dependency between the gradient descent iterates and the data.

In this paper, we describe a solution to tackle a common set of challenges in e-commerce, which arise from the fact that new products are continually being added to the catalogue. The challenges involve properly personalising the customer experience, forecasting demand and planning the product range. We argue that the foundational piece to solve all of these problems is having consistent and detailed information about each product, information that is rarely available or consistent given the multitude of suppliers and types of products. We describe in detail the architecture and methodology implemented at ASOS, one of the world's largest fashion e-commerce retailers, to tackle this problem. We then show how this quantitative understanding of the products can be leveraged to improve recommendations in a hybrid recommender system approach.

Sparse models for high-dimensional linear regression and machine learning have received substantial attention over the past two decades. Model selection, or determining which features or covariates are the best explanatory variables, is critical to the interpretability of a learned model. Much of the current literature assumes that covariates are only mildly correlated. However, in modern applications ranging from functional MRI to genome-wide association studies, covariates are highly correlated and do not exhibit key properties (such as the restricted eigenvalue condition, RIP, or other related assumptions). This paper considers a high-dimensional regression setting in which a graph governs both correlations among the covariates and the similarity among regression coefficients. Using side information about the strength of correlations among features, we form a graph with edge weights corresponding to pairwise covariances. This graph is used to define a graph total variation regularizer that promotes similar weights for highly correlated features. The graph structure encapsulated by this regularizer helps precondition correlated features to yield provably accurate estimates. Using graph-based regularizers to develop theoretical guarantees for highly-correlated covariates has not been previously examined. This paper shows how our proposed graph-based regularization yields mean-squared error guarantees for a broad range of covariance graph structures and correlation strengths which in many cases are optimal by imposing additional structure on $\beta^{\star}$ which encourages \emph{alignment} with the covariance graph. Our proposed approach outperforms other state-of-the-art methods for highly-correlated design in a variety of experiments on simulated and real fMRI data.

Domain adaptation (DA) is the task of classifying an unlabeled dataset (target) using a labeled dataset (source) from a related domain. The majority of successful DA methods try to directly match the distributions of the source and target data by transforming the feature space. Despite their success, state of the art methods based on this approach are either involved or unable to directly scale to data with many features. This article shows that domain adaptation can be successfully performed by using a very simple randomized expectation maximization (EM) method. We consider two instances of the method, which involve logistic regression and support vector machine, respectively. The underlying assumption of the proposed method is the existence of a good single linear classifier for both source and target domain. The potential limitations of this assumption are alleviated by the flexibility of the method, which can directly incorporate deep features extracted from a pre-trained deep neural network. The resulting algorithm is strikingly easy to implement and apply. We test its performance on 36 real-life adaptation tasks over text and image data with diverse characteristics. The method achieves state-of-the-art results, competitive with those of involved end-to-end deep transfer-learning methods.

We uncover a fairly general principle in online learning: If regret can be (approximately) expressed as a function of certain "sufficient statistics" for the data sequence, then there exists a special Burkholder function that 1) can be used algorithmically to achieve the regret bound and 2) only depends on these sufficient statistics, not the entire data sequence, so that the online strategy is only required to keep the sufficient statistics in memory. This characterization is achieved by bringing the full power of the Burkholder Method --- originally developed for certifying probabilistic martingale inequalities --- to bear on the online learning setting. To demonstrate the scope and effectiveness of the Burkholder method, we develop a novel online strategy for matrix prediction that attains a regret bound corresponding to the variance term in matrix concentration inequalities. We also present a linear-time/space prediction strategy for parameter free supervised learning with linear classes and general smooth norms.

In this paper, we introduce a powerful technique, Leave-One-Out, to the analysis of low-rank matrix completion problems. Using this technique, we develop a general approach for obtaining fine-grained, entry-wise bounds on iterative stochastic procedures. We demonstrate the power of this approach in analyzing two of the most important algorithms for matrix completion: the non-convex approach based on Singular Value Projection (SVP), and the convex relaxation approach based on nuclear norm minimization (NNM). In particular, we prove for the first time that the original form of SVP, without re-sampling or sample splitting, converges linearly in the infinity norm. We further apply our leave-one-out approach to an iterative procedure that arises in the analysis of the dual solutions of NNM. Our results show that NNM recovers the true $ d $-by-$ d $ rank-$ r $ matrix with $\mathcal{O}(\mu^2 r^3d \log d )$ observed entries, which has optimal dependence on the dimension and is independent of the condition number of the matrix. To the best of our knowledge, this is the first sample complexity result for a tractable matrix completion algorithm that satisfies these two properties simultaneously.

The central aim in this paper is to address variable selection questions in nonlinear and nonparametric regression. Motivated by statistical genetics, where nonlinear interactions are of particular interest, we introduce a novel, interpretable, and computationally efficient way to summarize the relative importance of predictor variables. Methodologically, we develop the "RelATive cEntrality" (RATE) measure to prioritize candidate genetic variants that are not just marginally important, but whose associations also stem from significant covarying relationships with other variants in the data. We illustrate RATE through Bayesian Gaussian process regression, but the methodological innovations apply to other nonlinear methods. It is known that nonlinear models often exhibit greater predictive accuracy than linear models, particularly for phenotypes generated by complex genetic architectures. With detailed simulations and an Arabidopsis thaliana QTL mapping study, we show that applying RATE enables an explanation for this improved performance.

Unsupervised Domain Adaptation (DA) is used to automatize the task of labeling data: an unlabeled dataset (target) is annotated using a labeled dataset (source) from a related domain. We cast domain adaptation as the problem of finding stable labels for target examples. A new definition of label stability is proposed, motivated by a generalization error bound for large margin linear classifiers: a target labeling is stable when, with high probability, a classifier trained on a random subsample of the target with that labeling yields the same labeling. We find stable labelings using a random walk on a directed graph with transition probabilities based on labeling stability. The majority vote of those labelings visited by the walk yields a stable label for each target example. The resulting domain adaptation algorithm is strikingly easy to implement and apply: It does not rely on data transformations, which are in general computational prohibitive in the presence of many input features, and does not need to access the source data, which is advantageous when data sharing is restricted. By acting on the original feature space, our method is able to take full advantage of deep features from external pre-trained neural networks, as demonstrated by the results of our experiments.

Recently, increasing attention has been directed to the study of the speech emotion recognition, in which global acoustic features of an utterance are mostly used to eliminate the content differences. However, the expression of speech emotion is a dynamic process, which is reflected through dynamic durations, energies, and some other prosodic information when one speaks. In this paper, a novel local dynamic pitch probability distribution feature, which is obtained by drawing the histogram, is proposed to improve the accuracy of speech emotion recognition. Compared with most of the previous works using global features, the proposed method takes advantage of the local dynamic information conveyed by the emotional speech. Several experiments on Berlin Database of Emotional Speech are conducted to verify the effectiveness of the proposed method. The experimental results demonstrate that the local dynamic information obtained with the proposed method is more effective for speech emotion recognition than the traditional global features.