In large scale machine learning, random sampling is a popular way to approximate datasets by a small representative subset of examples. In particular, sensitivity sampling is an intensely studied technique which provides provable guarantees on the quality of approximation, while reducing the number of examples to the product of the VC dimension $d$ and the total sensitivity $\mathfrak S$ in remarkably general settings. However, guarantees going beyond this general bound of $\mathfrak S d$ are known in perhaps only one setting, for $\ell_2$ subspace embeddings, despite intense study of sensitivity sampling in prior work. In this work, we show the first bounds for sensitivity sampling for $\ell_p$ subspace embeddings for $p > 2$ that improve over the general $\mathfrak S d$ bound, achieving a bound of roughly $\mathfrak S^{2-2/p}$ for $2

Inference and simulation in the context of high-dimensional dynamical systems remain computationally challenging problems. Some form of dimensionality reduction is required to make the problem tractable in general. In this paper, we propose a novel approximate Gaussian filtering and smoothing method which propagates low-rank approximations of the covariance matrices. This is accomplished by projecting the Lyapunov equations associated with the prediction step to a manifold of low-rank matrices, which are then solved by a recently developed, numerically stable, dynamical low-rank integrator. Meanwhile, the update steps are made tractable by noting that the covariance update only transforms the column space of the covariance matrix, which is low-rank by construction. The algorithm differentiates itself from existing ensemble-based approaches in that the low-rank approximations of the covariance matrices are deterministic, rather than stochastic. Crucially, this enables the method to reproduce the exact Kalman filter as the low-rank dimension approaches the true dimensionality of the problem. Our method reduces computational complexity from cubic (for the Kalman filter) to \emph{quadratic} in the state-space size in the worst-case, and can achieve \emph{linear} complexity if the state-space model satisfies certain criteria. Through a set of experiments in classical data-assimilation and spatio-temporal regression, we show that the proposed method consistently outperforms the ensemble-based methods in terms of error in the mean and covariance with respect to the exact Kalman filter. This comes at no additional cost in terms of asymptotic computational complexity.

An accurate and substantial dataset is essential for training a reliable and well-performing model. However, even manually annotated datasets contain label errors, not to mention automatically labeled ones. Previous methods for label denoising have primarily focused on detecting outliers and their permanent removal - a process that is likely to over- or underfilter the dataset. In this work, we propose AGRA: a new method for learning with noisy labels by using Adaptive GRAdient-based outlier removal. Instead of cleaning the dataset prior to model training, the dataset is dynamically adjusted during the training process. By comparing the aggregated gradient of a batch of samples and an individual example gradient, our method dynamically decides whether a corresponding example is helpful for the model at this point or is counter-productive and should be left out for the current update. Extensive evaluation on several datasets demonstrates AGRA's effectiveness, while a comprehensive results analysis supports our initial hypothesis: permanent hard outlier removal is not always what model benefits the most from.

We consider observations $(X,y)$ from single index models with unknown link function, Gaussian covariates and a regularized M-estimator $\hat\beta$ constructed from convex loss function and regularizer. In the regime where sample size $n$ and dimension $p$ are both increasing such that $p/n$ has a finite limit, the behavior of the empirical distribution of $\hat\beta$ and the predicted values $X\hat\beta$ has been previously characterized in a number of models: The empirical distributions are known to converge to proximal operators of the loss and penalty in a related Gaussian sequence model, which captures the interplay between ratio $p/n$, loss, regularization and the data generating process. This connection between$(\hat\beta,X\hat\beta)$ and the corresponding proximal operators require solving fixed-point equations that typically involve unobservable quantities such as the prior distribution on the index or the link function. This paper develops a different theory to describe the empirical distribution of $\hat\beta$ and $X\hat\beta$: Approximations of $(\hat\beta,X\hat\beta)$ in terms of proximal operators are provided that only involve observable adjustments. These proposed observable adjustments are data-driven, e.g., do not require prior knowledge of the index or the link function. These new adjustments yield confidence intervals for individual components of the index, as well as estimators of the correlation of $\hat\beta$ with the index. The interplay between loss, regularization and the model is thus captured in a data-driven manner, without solving the fixed-point equations studied in previous works. The results apply to both strongly convex regularizers and unregularized M-estimation. Simulations are provided for the square and logistic loss in single index models including logistic regression and 1-bit compressed sensing with 20\% corrupted bits.

This paper discusses pairing double/debiased machine learning (DDML) with stacking, a model averaging method for combining multiple candidate learners, to estimate structural parameters. We introduce two new stacking approaches for DDML: short-stacking exploits the cross-fitting step of DDML to substantially reduce the computational burden and pooled stacking enforces common stacking weights over cross-fitting folds. Using calibrated simulation studies and two applications estimating gender gaps in citations and wages, we show that DDML with stacking is more robust to partially unknown functional forms than common alternative approaches based on single pre-selected learners. We provide Stata and R software implementing our proposals.

To further reduce the computational effort in evaluating the approximate posterior, stochastic methods such as We introduce ensembles of stochastic neural networks to Monte Carlo dropout [47] and DropConnect [51] inference approximate the Bayesian posterior, combining stochastic have also been used extensively [13, 14, 40]. They benefit methods such as dropout with deep ensembles. The stochastic from computationally cheaper inference by virtue of sampling ensembles are formulated as families of distributions and stochastically from a single model. Formulated as a trained to approximate the Bayesian posterior with variational variational approximation to the posterior, dropout samples inference. We implement stochastic ensembles based from a family of parameter distributions where parameters on Monte Carlo dropout, DropConnect and a novel nonparametric can be randomly set to zero. Although this particular family version of dropout and evaluate them on a toy of distributions might seem unnatural [11], it turns out problem and CIFAR image classification. For both tasks, that the stochastic property can help to find more robust regions we test the quality of the posteriors directly against Hamiltonian of the parameter space, a fact well-known from a long Monte Carlo simulations. Our results show that history of using dropout as a regularization method.

We present a unified framework for deriving PAC-Bayesian generalization bounds. Unlike most previous literature on this topic, our bounds are anytime-valid (i.e., time-uniform), meaning that they hold at all stopping times, not only for a fixed sample size. Our approach combines four tools in the following order: (a) nonnegative supermartingales or reverse submartingales, (b) the method of mixtures, (c) the Donsker-Varadhan formula (or other convex duality principles), and (d) Ville's inequality. Our main result is a PAC-Bayes theorem which holds for a wide class of discrete stochastic processes. We show how this result implies time-uniform versions of well-known classical PAC-Bayes bounds, such as those of Seeger, McAllester, Maurer, and Catoni, in addition to many recent bounds. We also present several novel bounds. Our framework also enables us to relax traditional assumptions; in particular, we consider nonstationary loss functions and non-i.i.d. data. In sum, we unify the derivation of past bounds and ease the search for future bounds: one may simply check if our supermartingale or submartingale conditions are met and, if so, be guaranteed a (time-uniform) PAC-Bayes bound.

Economic models produce moment inequalities, which can be used to form tests of the true parameters. Confidence sets (CS) of the true parameters are derived by inverting these tests. However, they often lack analytical expressions, necessitating a grid search to obtain the CS numerically by retaining the grid points that pass the test. When the statistic is not asymptotically pivotal, constructing the critical value for each grid point in the parameter space adds to the computational burden. In this paper, we convert the computational issue into a classification problem by using a support vector machine (SVM) classifier. Its decision function provides a faster and more systematic way of dividing the parameter space into two regions: inside vs. outside of the confidence set. We label those points in the CS as 1 and those outside as -1. Researchers can train the SVM classifier on a grid of manageable size and use it to determine whether points on denser grids are in the CS or not. We establish certain conditions for the grid so that there is a tuning that allows us to asymptotically reproduce the test in the CS. This means that in the limit, a point is classified as belonging to the confidence set if and only if it is labeled as 1 by the SVM.

This review aims to contribute to the quest for artificial general intelligence by examining neuroscience and cognitive psychology methods for potential inspiration. Despite the impressive advancements achieved by deep learning models in various domains, they still have shortcomings in abstract reasoning and causal understanding. Such capabilities should be ultimately integrated into artificial intelligence systems in order to surpass data-driven limitations and support decision making in a way more similar to human intelligence. This work is a vertical review that attempts a wide-ranging exploration of brain function, spanning from lower-level biological neurons, spiking neural networks, and neuronal ensembles to higher-level concepts such as brain anatomy, vector symbolic architectures, cognitive and categorization models, and cognitive architectures. The hope is that these concepts may offer insights for solutions in artificial general intelligence.

Artificial intelligence (AI) has seen tremendous development in industry and academia. However, striking recent advances by industry have stunned the world, inviting a fresh perspective on the role of academic research in this field. Here, we characterize the impact and type of AI produced by both environments over the last 25 years and establish several patterns. We find that articles published by teams consisting exclusively of industry researchers tend to get greater attention, with a higher chance of being highly cited and citation-disruptive, and several times more likely to produce state-of-the-art models. In contrast, we find that exclusively academic teams publish the bulk of AI research and tend to produce higher novelty work, with single papers having several times higher likelihood of being unconventional and atypical. The respective impact-novelty advantages of industry and academia are robust to controls for subfield, team size, seniority, and prestige. We find that academic-industry collaborations struggle to replicate the novelty of academic teams and tend to look similar to industry teams. Together, our findings identify the unique and nearly irreplaceable contributions that both academia and industry make toward the healthy progress of AI.