Stochastic gradient descent (SGD) optimization methods are nowadays the method of choice for the training of deep neural networks (DNNs) in artificial intelligence systems. In practically relevant training problems, usually not the plain vanilla standard SGD method is the employed optimization scheme but instead suitably accelerated and adaptive SGD optimization methods are applied. As of today, maybe the most popular variant of such accelerated and adaptive SGD optimization methods is the famous Adam optimizer proposed by Kingma & Ba in 2014. Despite the popularity of the Adam optimizer in implementations, it remained an open problem of research to provide a convergence analysis for the Adam optimizer even in the situation of simple quadratic stochastic optimization problems where the objective function (the function one intends to minimize) is strongly convex. In this work we solve this problem by establishing optimal convergence rates for the Adam optimizer for a large class of stochastic optimization problems, in particular, covering simple quadratic stochastic optimization problems. The key ingredient of our convergence analysis is a new vector field function which we propose to refer to as the Adam vector field. This Adam vector field accurately describes the macroscopic behaviour of the Adam optimization process but differs from the negative gradient of the objective function (the function we intend to minimize) of the considered stochastic optimization problem. In particular, our convergence analysis reveals that the Adam optimizer does typically not converge to critical points of the objective function (zeros of the gradient of the objective function) of the considered optimization problem but converges with rates to zeros of this Adam vector field.

Adversarial attacks against deep learning models represent a major threat to the security and reliability of natural language processing (NLP) systems. In this paper, we propose a modification to the BERT-Attack framework, integrating Projected Gradient Descent (PGD) to enhance its effectiveness and robustness. The original BERT-Attack, designed for generating adversarial examples against BERT-based models, suffers from limitations such as a fixed perturbation budget and a lack of consideration for semantic similarity. The proposed approach in this work, PGD-BERT-Attack, addresses these limitations by leveraging PGD to iteratively generate adversarial examples while ensuring both imperceptibility and semantic similarity to the original input. Extensive experiments are conducted to evaluate the performance of PGD-BERT-Attack compared to the original BERT-Attack and other baseline methods. The results demonstrate that PGD-BERT-Attack achieves higher success rates in causing misclassification while maintaining low perceptual changes. Furthermore, PGD-BERT-Attack produces adversarial instances that exhibit greater semantic resemblance to the initial input, enhancing their applicability in real-world scenarios. Overall, the proposed modification offers a more effective and robust approach to adversarial attacks on BERT-based models, thus contributing to the advancement of defense against attacks on NLP systems.

Missing values are prevalent across various fields, posing challenges for training and deploying predictive models. In this context, imputation is a common practice, driven by the hope that accurate imputations will enhance predictions. However, recent theoretical and empirical studies indicate that simple constant imputation can be consistent and competitive. This empirical study aims at clarifying if and when investing in advanced imputation methods yields significantly better predictions. Relating imputation and predictive accuracies across combinations of imputation and predictive models on 20 datasets, we show that imputation accuracy matters less i) when using expressive models, ii) when incorporating missingness indicators as complementary inputs, iii) matters much more for generated linear outcomes than for real-data outcomes. Interestingly, we also show that the use of the missingness indicator is beneficial to the prediction performance, even in MCAR scenarios. Overall, on real-data with powerful models, improving imputation only has a minor effect on prediction performance. Thus, investing in better imputations for improved predictions often offers limited benefits.

The performance of sensor arrays in sensing and wireless communications improves with more elements, but this comes at the cost of increased energy consumption and hardware expense. This work addresses the challenge of selecting $k$ sensor elements from a set of $m$ to optimize a generic Quality-of-Service metric. Evaluating all $\binom{m}{k}$ possible sensor subsets is impractical, leading to prior solutions using convex relaxations, greedy algorithms, and supervised learning approaches. The current paper proposes a new framework that employs deep generative modeling, treating sensor selection as a deterministic Markov Decision Process where sensor subsets of size $k$ arise as terminal states. Generative Flow Networks (GFlowNets) are employed to model an action distribution conditioned on the state. Sampling actions from the aforementioned distribution ensures that the probability of arriving at a terminal state is proportional to the performance of the corresponding subset. Applied to a standard sensor selection scenario, the developed approach outperforms popular methods which are based on convex optimization and greedy algorithms. Finally, a multiobjective formulation of the proposed approach is adopted and applied on the sparse antenna array design for Integrated Sensing and Communication (ISAC) systems. The multiobjective variation is shown to perform well in managing the trade-off between radar and communication performance.

Our main focus is on the generalization bound, which serves as an upper limit for the generalization error. Our analysis delves into regression and classification tasks separately to ensure a thorough examination. We assume the target function is real-valued and Lipschitz continuous for regression tasks. We use the 2-norm and a root-mean-square-error (RMSE) variant to measure the disparities between predictions and actual values. In the case of classification tasks, we treat the target function as a one-hot classifier, representing a piece-wise constant function, and employ 0/1 loss for error measurement. Our analysis underscores the differing sample complexity required to achieve a concentration inequality of generalization bounds, highlighting the variation in learning efficiency for regression and classification tasks. Furthermore, we demonstrate that the generalization bounds for regression and classification functions are inversely proportional to a polynomial of the number of parameters in a network, with the degree depending on the hypothesis class and the network architecture. These findings emphasize the advantages of over-parameterized networks and elucidate the conditions for benign overfitting in such systems.

Bayesian Inference with Markov Chain Monte Carlo requires efficient computation of the likelihood function. In some scientific applications, the likelihood must be computed by numerically solving a partial differential equation, which can be prohibitively expensive. We demonstrate that some such problems can be made tractable by amortizing the computation with a surrogate likelihood function implemented by a neural network. We show that this has two additional benefits: reducing noise in the likelihood evaluations and providing fast gradient calculations. In experiments, the approach is applied to a model of heliospheric transport of galactic cosmic rays, where it enables efficient sampling from the posterior of latent parameters in the Parker equation.

The recent advance in neural rendering has enabled the ability to reconstruct high-quality 4D scenes using neural networks. Although 4D neural reconstruction is popular, registration for such representations remains a challenging task, especially for dynamic scene registration in surgical planning and simulation. In this paper, we propose a novel strategy for dynamic surgical neural scene registration. We first utilize 4D Gaussian Splatting to represent the surgical scene and capture both static and dynamic scenes effectively. Then, a spatial aware feature aggregation method, Spatially Weight Cluttering (SWC) is proposed to accurately align the feature between surgical scenes, enabling precise and realistic surgical simulations. Lastly, we present a novel strategy of deformable scene registration to register two dynamic scenes. By incorporating both spatial and temporal information for correspondence matching, our approach achieves superior performance compared to existing registration methods for implicit neural representation. The proposed method has the potential to improve surgical planning and training, ultimately leading to better patient outcomes.

Long-term time series forecasting in centralized environments poses unique challenges regarding data privacy, communication overhead, and scalability. To address these challenges, we propose FedTime, a federated large language model (LLM) tailored for long-range time series prediction. Specifically, we introduce a federated pre-trained LLM with fine-tuning and alignment strategies. Prior to the learning process, we employ K-means clustering to partition edge devices or clients into distinct clusters, thereby facilitating more focused model training. We also incorporate channel independence and patching to better preserve local semantic information, ensuring that important contextual details are retained while minimizing the risk of information loss. We demonstrate the effectiveness of our FedTime model through extensive experiments on various real-world forecasting benchmarks, showcasing substantial improvements over recent approaches. In addition, we demonstrate the efficiency of FedTime in streamlining resource usage, resulting in reduced communication overhead.

Gaussian graphical models can be used to extract conditional dependencies between the features of the dataset. This is often done by making an independence assumption about the samples, but this assumption is rarely satisfied in reality. However, state-of-the-art approaches that avoid this assumption are not scalable, with $O(n^3)$ runtime and $O(n^2)$ space complexity. In this paper, we introduce a method that has $O(n^2)$ runtime and $O(n)$ space complexity, without assuming independence. We validate our model on both synthetic and real-world datasets, showing that our method's accuracy is comparable to that of prior work We demonstrate that our approach can be used on unprecedentedly large datasets, such as a real-world 1,000,000-cell scRNA-seq dataset; this was impossible with previous approaches. Our method maintains the flexibility of prior work, such as the ability to handle multi-modal tensor-variate datasets and the ability to work with data of arbitrary marginal distributions. An additional advantage of our method is that, unlike prior work, our hyperparameters are easily interpretable.

In applications related to big data and service computing, dynamic connections tend to be encountered, especially the dynamic data of user-perspective quality of service (QoS) in Web services. They are transformed into high-dimensional and incomplete (HDI) tensors which include abundant temporal pattern information. Latent factorization of tensors (LFT) is an extremely efficient and typical approach for extracting such patterns from an HDI tensor. However, current LFT models require the QoS data to be maintained in a central place (e.g., a central server), which is impossible for increasingly privacy-sensitive users. To address this problem, this article creatively designs a federated learning based on latent factorization of tensors (FL-LFT). It builds a data-density -oriented federated learning model to enable isolated users to collaboratively train a global LFT model while protecting user's privacy. Extensive experiments on a QoS dataset collected from the real world verify that FL-LFT shows a remarkable increase in prediction accuracy when compared to state-of-the-art federated learning (FL) approaches.