The study of generalization properties of stochastic optimization algorithms has been at the heart of contemporary machine learning research. While in the more classical frameworks studies largely focused on the learning problem (e.g., Alon et al., 1997; Blumer et al., 1989), in the past decade it has become clear that in modern scenarios the particular algorithm used to learn the model plays a vital role in its generalization performance. As a prominent example, heavily over-parameterized deep neural networks trained by first order methods output models that generalize well, despite the fact that an arbitrarily chosen Empirical Risk Minimizer (ERM) may perform poorly (Zhang et al., 2017; Neyshabur et al., 2014, 2017). The present paper aims at understanding the generalization behavior of gradient methods, specifically in connection with the problem dimension, in the fundamental Stochastic Convex Optimization (SCO) learning setup; a well studied, theoretical framework widely used to study stochastic optimization algorithms. The seminal work of Shalev-Shwartz et al. (2010) was the first to show that uniform convergence, the canonical condition for generalization in statistical learning (e.g., Vapnik, 1971; Bartlett and Mendelson, 2002) may not hold in high-dimensional SCO: they demonstrated learning problems where there exist certain ERMs that overfit the training data (i.e., exhibit large population risk), while models produced by e.g., Stochastic Gradient Descent (SGD) or regularized empirical risk minimization generalize well. The construction presented by Shalev-Shwartz et al. (2010), however, featured a learning problem with dimension exponential in the number of training

It is necessary to develop efficient DNNs deployed on edge devices with limited computation resources. However, the compressed networks often execute new tasks in the target domain, which is different from the source domain where the original network is trained. It is important to investigate the robustness of compressed networks in two types of data distribution shifts: domain shifts and adversarial perturbations. In this study, we discover that compressed models are less robust to distribution shifts than their original networks. Interestingly, larger networks are more vulnerable to losing robustness than smaller ones, even when they are compressed to a similar size as the smaller networks. Furthermore, compact networks obtained by knowledge distillation are much more robust to distribution shifts than pruned networks. Finally, post-training quantization is a reliable method for achieving significant robustness to distribution shifts, and it outperforms both pruned and distilled models in terms of robustness.

The advancement of machine learning and symbolic approaches have underscored their strengths and weaknesses in Natural Language Processing (NLP). While machine learning approaches are powerful in identifying patterns in data, they often fall short in learning commonsense and the factual knowledge required for the NLP tasks. Meanwhile, the symbolic methods excel in representing knowledge-rich data. However, they struggle to adapt dynamic data and generalize the knowledge. Bridging these two paradigms through hybrid approaches enables the alleviation of weaknesses in both while preserving their strengths. Recent studies extol the virtues of this union, showcasing promising results in a wide range of NLP tasks. In this paper, we present an overview of hybrid approaches used for NLP. Specifically, we delve into the state-of-the-art hybrid approaches used for a broad spectrum of NLP tasks requiring natural language understanding, generation, and reasoning. Furthermore, we discuss the existing resources available for hybrid approaches for NLP along with the challenges, offering a roadmap for future directions.

This paper introduces the RUMBoost model, a novel discrete choice modelling approach that combines the interpretability and behavioural robustness of Random Utility Models (RUMs) with the generalisation and predictive ability of deep learning methods. We obtain the full functional form of non-linear utility specifications by replacing each linear parameter in the utility functions of a RUM with an ensemble of gradient boosted regression trees. This enables piece-wise constant utility values to be imputed for all alternatives directly from the data for any possible combination of input variables. We introduce additional constraints on the ensembles to ensure three crucial features of the utility specifications: (i) dependency of the utilities of each alternative on only the attributes of that alternative, (ii) monotonicity of marginal utilities, and (iii) an intrinsically interpretable functional form, where the exact response of the model is known throughout the entire input space. Furthermore, we introduce an optimisation-based smoothing technique that replaces the piece-wise constant utility values of alternative attributes with monotonic piece-wise cubic splines to identify non-linear parameters with defined gradient. We demonstrate the potential of the RUMBoost model compared to various ML and Random Utility benchmark models for revealed preference mode choice data from London. The results highlight the great predictive performance and the direct interpretability of our proposed approach. Furthermore, the smoothed attribute utility functions allow for the calculation of various behavioural indicators and marginal utilities. Finally, we demonstrate the flexibility of our methodology by showing how the RUMBoost model can be extended to complex model specifications, including attribute interactions, correlation within alternative error terms and heterogeneity within the population.

Accurate crop yield prediction is of utmost importance for informed decision-making in agriculture, aiding farmers, and industry stakeholders. However, this task is complex and depends on multiple factors, such as environmental conditions, soil properties, and management practices. Combining heterogeneous data views poses a fusion challenge, like identifying the view-specific contribution to the predictive task. We present a novel multi-view learning approach to predict crop yield for different crops (soybean, wheat, rapeseed) and regions (Argentina, Uruguay, and Germany). Our multi-view input data includes multi-spectral optical images from Sentinel-2 satellites and weather data as dynamic features during the crop growing season, complemented by static features like soil properties and topographic information. To effectively fuse the data, we introduce a Multi-view Gated Fusion (MVGF) model, comprising dedicated view-encoders and a Gated Unit (GU) module. The view-encoders handle the heterogeneity of data sources with varying temporal resolutions by learning a view-specific representation. These representations are adaptively fused via a weighted sum. The fusion weights are computed for each sample by the GU using a concatenation of the view-representations. The MVGF model is trained at sub-field level with 10 m resolution pixels. Our evaluations show that the MVGF outperforms conventional models on the same task, achieving the best results by incorporating all the data sources, unlike the usual fusion results in the literature. For Argentina, the MVGF model achieves an R2 value of 0.68 at sub-field yield prediction, while at field level evaluation (comparing field averages), it reaches around 0.80 across different countries. The GU module learned different weights based on the country and crop-type, aligning with the variable significance of each data source to the prediction task.

Traffic big data can be generally divided into two categories. The first type is Eulerian observations measured with fixed roadside sensors (e.g., loop detectors, traffic cameras, radars, etc.) installed on a small set of road links. This type of data is typically owned by municipal authorities (MAs). The second type is Lagrangian observations obtained from vehicle trajectory data (e.g., real-time vehicle locations and speeds), typically possessed by mobility providers (MPs) with large fleets, such as ride-hailing companies and public transport operators. Although these data can be useful for ITS applications, they provide only an incomplete picture of transportation systems due to the partial spatial coverage of the road network (in terms of Eulerian observations) and partial penetration rate over the vehicle population (in terms of Lagrangian observations). Traffic state estimation (TSE) is an important research topic that leverages these partially observed traffic data to infer key traffic state variables (e.g., flow, density, speed) on road segments.

Sparse identification of differential equations aims to compute the analytic expressions from the observed data explicitly. However, there exist two primary challenges. Firstly, it exhibits sensitivity to the noise in the observed data, particularly for the derivatives computations. Secondly, existing literature predominantly concentrates on single-fidelity (SF) data, which imposes limitations on its applicability due to the computational cost. In this paper, we present two novel approaches to address these problems from the view of uncertainty quantification. We construct a surrogate model employing the Gaussian process regression (GPR) to mitigate the effect of noise in the observed data, quantify its uncertainty, and ultimately recover the equations accurately. Subsequently, we exploit the multi-fidelity Gaussian processes (MFGP) to address scenarios involving multi-fidelity (MF), sparse, and noisy observed data. We demonstrate the robustness and effectiveness of our methodologies through several numerical experiments.

The safe integration of machine learning modules in decision-making processes hinges on their ability to quantify uncertainty. A popular technique to achieve this goal is conformal prediction (CP), which transforms an arbitrary base predictor into a set predictor with coverage guarantees. While CP certifies the predicted set to contain the target quantity with a user-defined tolerance, it does not provide control over the average size of the predicted sets, i.e., over the informativeness of the prediction. In this work, a theoretical connection is established between the generalization properties of the base predictor and the informativeness of the resulting CP prediction sets. To this end, an upper bound is derived on the expected size of the CP set predictor that builds on generalization error bounds for the base predictor. The derived upper bound provides insights into the dependence of the average size of the CP set predictor on the amount of calibration data, the target reliability, and the generalization performance of the base predictor. The theoretical insights are validated using simple numerical regression and classification tasks.

Most existing graph neural networks (GNNs) are limited to undirected graphs, whose restricted scope of the captured relational information hinders their expressive capabilities and deployments in real-world scenarios. Compared with undirected graphs, directed graphs (digraphs) fit the demand for modeling more complex topological systems by capturing more intricate relationships between nodes, such as formulating transportation and financial networks. While some directed GNNs have been introduced, their inspiration mainly comes from deep learning architectures, which lead to redundant complexity and computation, making them inapplicable to large-scale databases. To address these issues, we propose LightDiC, a scalable variant of the digraph convolution based on the magnetic Laplacian. Since topology-related computations are conducted solely during offline pre-processing, LightDiC achieves exceptional scalability, enabling downstream predictions to be trained separately without incurring recursive computational costs. Theoretical analysis shows that LightDiC utilizes directed information to achieve message passing based on the complex field, which corresponds to the proximal gradient descent process of the Dirichlet energy optimization function from the perspective of digraph signal denoising, ensuring its expressiveness. Experimental results demonstrate that LightDiC performs comparably well or even outperforms other SOTA methods in various downstream tasks, with fewer learnable parameters and higher training efficiency. Notably, LightDiC is the first DiGNN to provide satisfactory results in the most representative large-scale database (ogbn-papers100M).

With the rapid advancement of AI applications, the growing needs for data privacy and model robustness have highlighted the importance of machine unlearning, especially in thriving graph-based scenarios. However, most existing graph unlearning strategies primarily rely on well-designed architectures or manual process, rendering them less user-friendly and posing challenges in terms of deployment efficiency. Furthermore, striking a balance between unlearning performance and framework generalization is also a pivotal concern. To address the above issues, we propose \underline{\textbf{M}}utual \underline{\textbf{E}}volution \underline{\textbf{G}}raph \underline{\textbf{U}}nlearning (MEGU), a new mutual evolution paradigm that simultaneously evolves the predictive and unlearning capacities of graph unlearning. By incorporating aforementioned two components, MEGU ensures complementary optimization in a unified training framework that aligns with the prediction and unlearning requirements. Extensive experiments on 9 graph benchmark datasets demonstrate the superior performance of MEGU in addressing unlearning requirements at the feature, node, and edge levels. Specifically, MEGU achieves average performance improvements of 2.7\%, 2.5\%, and 3.2\% across these three levels of unlearning tasks when compared to state-of-the-art baselines. Furthermore, MEGU exhibits satisfactory training efficiency, reducing time and space overhead by an average of 159.8x and 9.6x, respectively, in comparison to retraining GNN from scratch.