Among generalized additive models, additive Mat\'ern Gaussian Processes (GPs) are one of the most popular for scalable high-dimensional problems. Thanks to their additive structure and stochastic differential equation representation, back-fitting-based algorithms can reduce the time complexity of computing the posterior mean from $O(n^3)$ to $O(n\log n)$ time where $n$ is the data size. However, generalizing these algorithms to efficiently compute the posterior variance and maximum log-likelihood remains an open problem. In this study, we demonstrate that for Additive Mat\'ern GPs, not only the posterior mean, but also the posterior variance, log-likelihood, and gradient of these three functions can be represented by formulas involving only sparse matrices and sparse vectors. We show how to use these sparse formulas to generalize back-fitting-based algorithms to efficiently compute the posterior mean, posterior variance, log-likelihood, and gradient of these three functions for additive GPs, all in $O(n \log n)$ time. We apply our algorithms to Bayesian optimization and propose efficient algorithms for posterior updates, hyperparameters learning, and computations of the acquisition function and its gradient in Bayesian optimization. Given the posterior, our algorithms significantly reduce the time complexity of computing the acquisition function and its gradient from $O(n^2)$ to $O(\log n)$ for general learning rate, and even to $O(1)$ for small learning rate.
相關內容
We consider linear random coefficient regression models, where the regressors are allowed to have a finite support. First, we investigate identifiability, and show that the means and the variances and covariances of the random coefficients are identified from the first two conditional moments of the response given the covariates if the support of the covariates, excluding the intercept, contains a Cartesian product with at least three points in each coordinate. We also discuss ientification of higher-order mixed moments, as well as partial identification in the presence of a binary regressor. Next we show the variable selection consistency of the adaptive LASSO for the variances and covariances of the random coefficients in finite and moderately high dimensions. This implies that the estimated covariance matrix will actually be positive semidefinite and hence a valid covariance matrix, in contrast to the estimate arising from a simple least squares fit. We illustrate the proposed method in a simulation study.
We study a variant of the widely popular, fast and often used family of community detection procedures referred to as label propagation algorithms. These mechanisms also exhibit many parallels with models of opinion exchange dynamics and consensus mechanisms in distributed computing. Initially, given a network, each vertex starts with a random label in the interval $[0,1]$. Then, in each round of the algorithm, every vertex switches its label to the majority label in its neighborhood (including its own label). At the first round, ties are broken towards smaller labels, while at each of the next rounds, ties are broken uniformly at random. We investigate the performance of this algorithm on the binomial random graph $\mathcal{G}(n,p)$. We show that for $np \ge n^{5/8+\varepsilon}$, the algorithm terminates with a single label a.a.s. (which was previously known only for $np\ge n^{3/4+\varepsilon}$). Moreover, we show that if $np\gg n^{2/3}$, a.a.s.\ this label is the smallest one, whereas if $n^{5/8+\varepsilon}\le np\ll n^{2/3}$, the surviving label is a.a.s. not the smallest one.
We consider finding flat, local minimizers by adding average weight perturbations. Given a nonconvex function $f: \mathbb{R}^d \rightarrow \mathbb{R}$ and a $d$-dimensional distribution $\mathcal{P}$ which is symmetric at zero, we perturb the weight of $f$ and define $F(W) = \mathbb{E}[f({W + U})]$, where $U$ is a random sample from $\mathcal{P}$. This injection induces regularization through the Hessian trace of $f$ for small, isotropic Gaussian perturbations. Thus, the weight-perturbed function biases to minimizers with low Hessian trace. Several prior works have studied settings related to this weight-perturbed function by designing algorithms to improve generalization. Still, convergence rates are not known for finding minima under the average perturbations of the function $F$. This paper considers an SGD-like algorithm that injects random noise before computing gradients while leveraging the symmetry of $\mathcal{P}$ to reduce variance. We then provide a rigorous analysis, showing matching upper and lower bounds of our algorithm for finding an approximate first-order stationary point of $F$ when the gradient of $f$ is Lipschitz-continuous. We empirically validate our algorithm for several image classification tasks with various architectures. Compared to sharpness-aware minimization, we note a 12.6% and 7.8% drop in the Hessian trace and top eigenvalue of the found minima, respectively, averaged over eight datasets. Ablation studies validate the benefit of the design of our algorithm.
In this work we connect two notions: That of the nonparametric mode of a probability measure, defined by asymptotic small ball probabilities, and that of the Onsager-Machlup functional, a generalized density also defined via asymptotic small ball probabilities. We show that in a separable Hilbert space setting and under mild conditions on the likelihood, modes of a Bayesian posterior distribution based upon a Gaussian prior exist and agree with the minimizers of its Onsager-Machlup functional and thus also with weak posterior modes. We apply this result to inverse problems and derive conditions on the forward mapping under which this variational characterization of posterior modes holds. Our results show rigorously that in the limit case of infinite-dimensional data corrupted by additive Gaussian or Laplacian noise, nonparametric maximum a posteriori estimation is equivalent to Tikhonov-Phillips regularization. In comparison with the work of Dashti, Law, Stuart, and Voss (2013), the assumptions on the likelihood are relaxed so that they cover in particular the important case of white Gaussian process noise. We illustrate our results by applying them to a severely ill-posed linear problem with Laplacian noise, where we express the maximum a posteriori estimator analytically and study its rate of convergence in the small noise limit.
We propose regression models for curve-valued responses in two or more dimensions, where only the image but not the parametrization of the curves is of interest. Examples of such data are handwritten letters, movement paths or outlines of objects. In the square-root-velocity framework, a parametrization invariant distance for curves is obtained as the quotient space metric with respect to the action of re-parametrization, which is by isometries. With this special case in mind, we discuss the generalization of 'linear' regression to quotient metric spaces more generally, before illustrating the usefulness of our approach for curves modulo re-parametrization. We address the issue of sparsely or irregularly sampled curves by using splines for modeling smooth conditional mean curves. We test this model in simulations and apply it to human hippocampal outlines, obtained from Magnetic Resonance Imaging scans. Here we model how the shape of the irregularly sampled hippocampus is related to age, Alzheimer's disease and sex.
Privacy-utility tradeoff remains as one of the fundamental issues of differentially private machine learning. This paper introduces a geometrically inspired kernel-based approach to mitigate the accuracy-loss issue in classification. In this approach, a representation of the affine hull of given data points is learned in Reproducing Kernel Hilbert Spaces (RKHS). This leads to a novel distance measure that hides privacy-sensitive information about individual data points and improves the privacy-utility tradeoff via significantly reducing the risk of membership inference attacks. The effectiveness of the approach is demonstrated through experiments on MNIST dataset, Freiburg groceries dataset, and a real biomedical dataset. It is verified that the approach remains computationally practical. The application of the approach to federated learning is considered and it is observed that the accuracy-loss due to data being distributed is either marginal or not significantly high.
Angluin's L$^*$ algorithm learns the minimal deterministic finite automaton (DFA) of a regular language using membership and equivalence queries. Its probabilistic approximatively correct (PAC) version substitutes an equivalence query by numerous random membership queries to get a high level confidence to the answer. Thus it can be applied to any kind of device and may be viewed as an algorithm for synthesizing an automaton abstracting the behavior of the device based on observations. Here we are interested on how Angluin's PAC learning algorithm behaves for devices which are obtained from a DFA by introducing some noise. More precisely we study whether Angluin's algorithm reduces the noise and produces a DFA closer to the original one than the noisy device. We propose several ways to introduce the noise: (1) the noisy device inverts the classification of words w.r.t. the DFA with a small probability, (2) the noisy device modifies with a small probability the letters of the word before asking its classification w.r.t. the DFA, (3) the noisy device combines the classification of a word w.r.t. the DFA and its classification w.r.t. a counter automaton, and (4) the noisy DFA is obtained by a random process from two DFA such that the language of the first one is included in the second one. Then when a word is accepted (resp. rejected) by the first (resp. second) one, it is also accepted (resp. rejected) and in the remaining cases, it is accepted with probability 0.5. Our main experimental contributions consist in showing that: (1) Angluin's algorithm behaves well whenever the noisy device is produced by a random process, (2) but poorly with a structured noise, and, that (3) is able to eliminate pathological behaviours specified in a regular way. Theoretically, we show that randomness almost surely yields systems with non-recursively enumerable languages.
In online advertisement, ad campaigns are sequentially displayed to users. Both users and campaigns have inherent features, and the former is eligible to the latter if they are ``similar enough''. We model these interactions as a bipartite geometric random graph: the features of the $2N$ vertices ($N$ users and $N$ campaigns) are drawn independently in a metric space and an edge is present between a campaign and a user node if the distance between their features is smaller than $c/N$, where $c>0$ is the parameter of the model. Our contributions are two-fold. In the one-dimensional case, with uniform distribution over the segment $[0,1]$, we derive the size of the optimal offline matching in these bipartite random geometric graphs, and we build an algorithm achieving it (as a benchmark), and analyze precisely its performance. We then turn to the online setting where one side of the graph is known at the beginning while the other part is revealed sequentially. We study the number of matches of the online algorithm closest, which matches any incoming point to its closest available neighbor. We show that its performances can be compared to its fluid limit, completely described as the solution of an explicit PDE. From the latter, we can compute the competitive ratio of closest.
Click-Through Rate (CTR) prediction is a pivotal task in product and content recommendation, where learning effective feature embeddings is of great significance. However, traditional methods typically learn fixed feature representations without dynamically refining feature representations according to the context information, leading to suboptimal performance. Some recent approaches attempt to address this issue by learning bit-wise weights or augmented embeddings for feature representations, but suffer from uninformative or redundant features in the context. To tackle this problem, inspired by the Global Workspace Theory in conscious processing, which posits that only a specific subset of the product features are pertinent while the rest can be noisy and even detrimental to human-click behaviors, we propose a CTR model that enables Dynamic Embedding Learning with Truncated Conscious Attention for CTR prediction, termed DELTA. DELTA contains two key components: (I) conscious truncation module (CTM), which utilizes curriculum learning to apply adaptive truncation on attention weights to select the most critical feature in the context; (II) explicit embedding optimization (EEO), which applies an auxiliary task during training that directly and independently propagates the gradient from the loss layer to the embedding layer, thereby optimizing the embedding explicitly via linear feature crossing. Extensive experiments on five challenging CTR datasets demonstrate that DELTA achieves new state-of-art performance among current CTR methods.
Standard contrastive learning approaches usually require a large number of negatives for effective unsupervised learning and often exhibit slow convergence. We suspect this behavior is due to the suboptimal selection of negatives used for offering contrast to the positives. We counter this difficulty by taking inspiration from support vector machines (SVMs) to present max-margin contrastive learning (MMCL). Our approach selects negatives as the sparse support vectors obtained via a quadratic optimization problem, and contrastiveness is enforced by maximizing the decision margin. As SVM optimization can be computationally demanding, especially in an end-to-end setting, we present simplifications that alleviate the computational burden. We validate our approach on standard vision benchmark datasets, demonstrating better performance in unsupervised representation learning over state-of-the-art, while having better empirical convergence properties.
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191