Identification of high-risk driving situations is generally approached through collision risk estimation or accident pattern recognition. In this work, we approach the problem from the perspective of subjective risk. We operationalize subjective risk assessment by predicting driver behavior changes and identifying the cause of changes. To this end, we introduce a new task called driver-centric risk object identification (DROID), which uses egocentric video to identify object(s) influencing a driver's behavior, given only the driver's response as the supervision signal. We formulate the task as a cause-effect problem and present a novel two-stage DROID framework, taking inspiration from models of situation awareness and causal inference. A subset of data constructed from the Honda Research Institute Driving Dataset (HDD) is used to evaluate DROID. We demonstrate state-of-the-art DROID performance, even compared with strong baseline models using this dataset. Additionally, we conduct extensive ablative studies to justify our design choices. Moreover, we demonstrate the applicability of DROID for risk assessment.

Stochastic optimization is a fundamental area with extensive applications in many domains, ranging from machine learning to algorithm design and beyond. The design and analysis of iterative methods for stochastic optimization has been the focus of a long line of work, leading to a rich understanding of the convergence of paradigmatic iterative methods such as stochastic gradient descent, mirror descent, and accelerated methods for both convex and non-convex optimization. However, most of these works only establish convergence guarantees that hold only in expectation. Although very meaningful, these results do not fully capture the convergence behaviors of the algorithms when we perform only a small number of runs of the algorithm, as it is typical in modern machine learning applications where there are significant computational and statistical costs associated with performing multiple runs of the algorithm (Harvey et al., 2019; Madden et al., 2020; Davis et al., 2021). Thus, an important direction is to establish convergence guarantees for a single run of the algorithm that hold not only in expectation but also with high probability. Compared to the guarantees that hold in expectation, high probability guarantees are significantly harder to obtain and they hold in more limited settings with stronger assumptions on the problem settings and the stochastic noise distribution. Most existing works that establish high probability guarantees focus on the setting where the length of the stochastic noise follows a light-tail (sub-Gaussian) distribution (Juditsky et al., 2011; Lan, 2012, 2020; Li and Orabona, 2020; Madden et al., 2020; Kavis et al., 2021). Recent works also study the more challenging heavy-tail setting, notably under a bounded variance (Nazin et al., 2019; Gorbunov et al., 2020; Cutkosky and Mehta, 2021) or bounded p-moment assumption (Cutkosky and Mehta, 2021) on the length of the stochastic noise. Both settings are highly relevant in practice: Zhang et al. (2020) empirically studied the noise distribution for two common tasks, training a

Temporal distributional shifts, with underlying dynamics changing over time, frequently occur in real-world time series, and pose a fundamental challenge for deep neural networks (DNNs). In this paper, we propose a novel deep sequence model based on the Koopman theory for time series forecasting: Koopman Neural Forecaster (KNF) that leverages DNNs to learn the linear Koopman space and the coefficients of chosen measurement functions. KNF imposes appropriate inductive biases for improved robustness against distributional shifts, employing both a global operator to learn shared characteristics, and a local operator to capture changing dynamics, as well as a specially-designed feedback loop to continuously update the learnt operators over time for rapidly varying behaviors. We demonstrate that KNF achieves the superior performance compared to the alternatives, on multiple time series datasets that are shown to suffer from distribution shifts. Temporal distribution shifts frequently occur in real-world time-series applications, from forecasting stock prices to detecting and monitoring sensory measures, to predicting fashion trend based sales. Such distribution shifts over time may due to the data being generated in a highly-dynamic and non-stationary environment, abrupt changes that are difficult to predict, or constantly evolving trends in the underlying data distribution (Gama et al., 2014). Temporal distribution shifts pose a fundamental challenge for time-series forecasting (Kuznetsov & Mohri, 2020). There are two scenarios of distribution shifts. When the distribution shifts only occur between the training and test domains, meta learning and transfer learning approaches (Jin et al., 2021; Oreshkin et al., 2021) have been developed.

The sheer size of modern neural networks makes model serving a serious computational challenge. A popular class of compression techniques overcomes this challenge by pruning or sparsifying the weights of pretrained networks. While useful, these techniques often face serious tradeoffs between computational requirements and compression quality. In this work, we propose a novel optimization-based pruning framework that considers the combined effect of pruning (and updating) multiple weights subject to a sparsity constraint. Our approach, CHITA, extends the classical Optimal Brain Surgeon framework and results in significant improvements in speed, memory, and performance over existing optimization-based approaches for network pruning. CHITA's main workhorse performs combinatorial optimization updates on a memory-friendly representation of local quadratic approximation(s) of the loss function. On a standard benchmark of pretrained models and datasets, CHITA leads to significantly better sparsity-accuracy tradeoffs than competing methods. For example, for MLPNet with only 2% of the weights retained, our approach improves the accuracy by 63% relative to the state of the art. Furthermore, when used in conjunction with fine-tuning SGD steps, our method achieves significant accuracy gains over the state-of-the-art approaches.

Many problems arising in control require the determination of a mathematical model of the application. This has often to be performed starting from input-output data, leading to a task known as system identification in the engineering literature. One emerging topic in this field is estimation of networks consisting of several interconnected dynamic systems. We consider the linear setting assuming that system outputs are the result of many correlated inputs, hence making system identification severely ill-conditioned. This is a scenario often encountered when modeling complex cybernetics systems composed by many sub-units with feedback and algebraic loops. We develop a strategy cast in a Bayesian regularization framework where any impulse response is seen as realization of a zero-mean Gaussian process. Any covariance is defined by the so called stable spline kernel which includes information on smooth exponential decay. We design a novel Markov chain Monte Carlo scheme able to reconstruct the impulse responses posterior by efficiently dealing with collinearity. Our scheme relies on a variation of the Gibbs sampling technique: beyond considering blocks forming a partition of the parameter space, some other (overlapping) blocks are also updated on the basis of the level of collinearity of the system inputs. Theoretical properties of the algorithm are studied obtaining its convergence rate. Numerical experiments are included using systems containing hundreds of impulse responses and highly correlated inputs.

In computer vision, it is often observed that formulating regression problems as a classification task yields better performance. We investigate this curious phenomenon and provide a derivation to show that classification, with the crossentropy loss, outperforms regression with a mean squared error loss in its ability to learn high-entropy feature representations. Based on the analysis, we propose an ordinal entropy regularizer to encourage higher-entropy feature spaces while maintaining ordinal relationships to improve the performance of regression tasks. Experiments on synthetic and real-world regression tasks demonstrate the importance and benefits of increasing entropy for regression. Classification and regression are two fundamental tasks of machine learning. The choice between the two usually depends on the categorical or continuous nature of the target output.

In this paper, we study the generalization performance of global minima for implementing empirical risk minimization (ERM) on over-parameterized deep ReLU nets. Using a novel deepening scheme for deep ReLU nets, we rigorously prove that there exist perfect global minima achieving almost optimal generalization error bounds for numerous types of data under mild conditions. Since over-parameterization is crucial to guarantee that the global minima of ERM on deep ReLU nets can be realized by the widely used stochastic gradient descent (SGD) algorithm, our results indeed fill a gap between optimization and generalization.

Federated Learning (FL) is a distributed machine learning approach where multiple clients work together to solve a machine learning task. One of the key challenges in FL is the issue of partial participation, which occurs when a large number of clients are involved in the training process. The traditional method to address this problem is randomly selecting a subset of clients at each communication round. In our research, we propose a new technique and design a novel regularized client participation scheme. Under this scheme, each client joins the learning process every $R$ communication rounds, which we refer to as a meta epoch. We have found that this participation scheme leads to a reduction in the variance caused by client sampling. Combined with the popular FedAvg algorithm (McMahan et al., 2017), it results in superior rates under standard assumptions. For instance, the optimization term in our main convergence bound decreases linearly with the product of the number of communication rounds and the size of the local dataset of each client, and the statistical term scales with step size quadratically instead of linearly (the case for client sampling with replacement), leading to better convergence rate $\mathcal{O}\left(\frac{1}{T^2}\right)$ compared to $\mathcal{O}\left(\frac{1}{T}\right)$, where $T$ is the total number of communication rounds. Furthermore, our results permit arbitrary client availability as long as each client is available for training once per each meta epoch.

Real-world vision based applications require fine-grained classification for various area of interest like e-commerce, mobile applications, warehouse management, etc. where reducing the severity of mistakes and improving the classification accuracy is of utmost importance. This paper proposes a method to boost fine-grained classification through a hierarchical approach via learnable independent query embeddings. This is achieved through a classification network that uses coarse class predictions to improve the fine class accuracy in a stage-wise sequential manner. We exploit the idea of hierarchy to learn query embeddings that are scalable across all levels, thus making this a relevant approach even for extreme classification where we have a large number of classes. The query is initialized with a weighted Eigen image calculated from training samples to best represent and capture the variance of the object. We introduce transformer blocks to fuse intermediate layers at which query attention happens to enhance the spatial representation of feature maps at different scales. This multi-scale fusion helps improve the accuracy of small-size objects. We propose a two-fold approach for the unique representation of learnable queries. First, at each hierarchical level, we leverage cluster based loss that ensures maximum separation between inter-class query embeddings and helps learn a better (query) representation in higher dimensional spaces. Second, we fuse coarse level queries with finer level queries weighted by a learned scale factor. We additionally introduce a novel block called Cross Attention on Multi-level queries with Prior (CAMP) Block that helps reduce error propagation from coarse level to finer level, which is a common problem in all hierarchical classifiers. Our method is able to outperform the existing methods with an improvement of ~11% at the fine-grained classification.

Cross-domain recommendation has attracted increasing attention from industry and academia recently. However, most existing methods do not exploit the interest invariance between domains, which would yield sub-optimal solutions. In this paper, we propose a cross-domain recommendation method: Self-supervised Interest Transfer Network (SITN), which can effectively transfer invariant knowledge between domains via prototypical contrastive learning. Specifically, we perform two levels of cross-domain contrastive learning: 1) instance-to-instance contrastive learning, 2) instance-to-cluster contrastive learning. Not only that, we also take into account users' multi-granularity and multi-view interests. With this paradigm, SITN can explicitly learn the invariant knowledge of interest clusters between domains and accurately capture users' intents and preferences. We conducted extensive experiments on a public dataset and a large-scale industrial dataset collected from one of the world's leading e-commerce corporations. The experimental results indicate that SITN achieves significant improvements over state-of-the-art recommendation methods. Additionally, SITN has been deployed on a micro-video recommendation platform, and the online A/B testing results further demonstrate its practical value. Supplement is available at: https://github.com/fanqieCoffee/SITN-Supplement.