Generalization is a central theme in statistical learning theory. The recent renewed interest in kernel methods, especially in Kernel Ridge Regression (KRR), is largely due to the fact that deep neural network (DNN) training can be approximated using kernels under appropriate conditions Jacot et al. [2018], Arora et al. [2019], Bordelon et al. [2020], in which the test error is more tractable analytically and thus enjoys stronger theoretical guarantees. However, many prior results have been derived under conditions incompatible with practical settings. For instance Liang and Rakhlin [2020], Liu et al. [2021a], Mei et al. [2021], Misiakiewicz [2022] give asymptotic bounds on the KRR test error, which requires the input dimension d to tend to infinity. In reality, the input dimension of the data set and the target function is typically finite.

Large molecular representation models pre-trained on massive unlabeled data have shown great success in predicting molecular properties. However, these models may tend to overfit the fine-tuning data, resulting in over-confident predictions on test data that fall outside of the training distribution. To address this issue, uncertainty quantification (UQ) methods can be used to improve the models' calibration of predictions. Although many UQ approaches exist, not all of them lead to improved performance. While some studies have included UQ to improve molecular pre-trained models, the process of selecting suitable backbone and UQ methods for reliable molecular uncertainty estimation remains underexplored. To address this gap, we present MUBen, which evaluates different UQ methods for state-of-the-art backbone molecular representation models to investigate their capabilities. By fine-tuning various backbones using different molecular descriptors as inputs with UQ methods from different categories, we critically assess the influence of architectural decisions and training strategies. Our study offers insights for selecting UQ for backbone models, which can facilitate research on uncertainty-critical applications in fields such as materials science and drug discovery.

The division of one physical 5G communications infrastructure into several virtual network slices with distinct characteristics such as bandwidth, latency, reliability, security, and service quality is known as 5G network slicing. Each slice is a separate logical network that meets the requirements of specific services or use cases, such as virtual reality, gaming, autonomous vehicles, or industrial automation. The network slice can be adjusted dynamically to meet the changing demands of the service, resulting in a more cost-effective and efficient approach to delivering diverse services and applications over a shared infrastructure. This paper assesses various machine learning techniques, including the logistic regression model, linear discriminant model, k-nearest neighbor's model, decision tree model, random forest model, SVC BernoulliNB model, and GaussianNB model, to investigate the accuracy and precision of each model on detecting network slices. The report also gives an overview of 5G network slicing.

Fault Localization (FL) aims to automatically localize buggy lines of code, a key first step in many manual and automatic debugging tasks. Previous FL techniques assume the provision of input tests, and often require extensive program analysis, program instrumentation, or data preprocessing. Prior work on deep learning for APR struggles to learn from small datasets and produces limited results on real-world programs. Inspired by the ability of large language models (LLMs) of code to adapt to new tasks based on very few examples, we investigate the applicability of LLMs to line level fault localization. Specifically, we propose to overcome the left-to-right nature of LLMs by fine-tuning a small set of bidirectional adapter layers on top of the representations learned by LLMs to produce LLMAO, the first language model based fault localization approach that locates buggy lines of code without any test coverage information. We fine-tune LLMs with 350 million, 6 billion, and 16 billion parameters on small, manually curated corpora of buggy programs such as the Defects4J corpus. We observe that our technique achieves substantially more confidence in fault localization when built on the larger models, with bug localization performance scaling consistently with the LLM size. Our empirical evaluation shows that LLMAO improves the Top-1 results over the state-of-the-art machine learning fault localization (MLFL) baselines by 2.3%-54.4%, and Top-5 results by 14.4%-35.6%. LLMAO is also the first FL technique trained using a language model architecture that can detect security vulnerabilities down to the code line level.

Reinforcement learning often requires millions of examples to produce static, black-box models. In contrast, interactive task learning (ITL) emphasizes incremental knowledge acquisition from limited instruction provided by humans in modalities such as natural language. However, in practice, ITL systems often suffers from brittle, error-prone language parsing. Large language models (LLMs) are resistant to brittleness but are not interpretable and cannot learn incrementally. We present VAL, an ITL system with a new philosophy for LLM/symbolic integration. By using LLMs only for specific tasks -- such as predicate and argument selection -- within an algorithmic framework, VAL reaps the benefits of LLMs to support interactive learning of hierarchical task knowledge from natural language. Acquired knowledge is human interpretable and generalizes to support execution of novel tasks without additional training. We studied users' interactions with VAL in a video game setting, finding that most users could successfully teach VAL using language they felt was natural.

Timely and accurate assessment of hurricane-induced building damage is crucial for effective post-hurricane response and recovery efforts. Recently, remote sensing technologies provide large-scale optical or Interferometric Synthetic Aperture Radar (InSAR) imagery data immediately after a disastrous event, which can be readily used to conduct rapid building damage assessment. Compared to optical satellite imageries, the Synthetic Aperture Radar can penetrate cloud cover and provide more complete spatial coverage of damaged zones in various weather conditions. However, these InSAR imageries often contain highly noisy and mixed signals induced by co-occurring or co-located building damage, flood, flood/wind-induced vegetation changes, as well as anthropogenic activities, making it challenging to extract accurate building damage information. In this paper, we introduced an approach for rapid post-hurricane building damage detection from InSAR imagery. This approach encoded complex causal dependencies among wind, flood, building damage, and InSAR imagery using a holistic causal Bayesian network. Based on the causal Bayesian network, we further jointly inferred the large-scale unobserved building damage by fusing the information from InSAR imagery with prior physical models of flood and wind, without the need for ground truth labels. Furthermore, we validated our estimation results in a real-world devastating hurricane -- the 2022 Hurricane Ian. We gathered and annotated building damage ground truth data in Lee County, Florida, and compared the introduced method's estimation results with the ground truth and benchmarked it against state-of-the-art models to assess the effectiveness of our proposed method. Results show that our method achieves rapid and accurate detection of building damage, with significantly reduced processing time compared to traditional manual inspection methods.

Training AI models that generalize across tasks and domains has long been among the open problems driving AI research. The emergence of Foundation Models made it easier to obtain expert models for a given task, but the heterogeneity of data that may be encountered at test time often means that any single expert is insufficient. We consider the Fusion of Experts (FoE) problem of fusing outputs of expert models with complementary knowledge of the data distribution and formulate it as an instance of supervised learning. Our method is applicable to both discriminative and generative tasks and leads to significant performance improvements in image and text classification, text summarization, multiple-choice QA, and automatic evaluation of generated text. We also extend our method to the "frugal" setting where it is desired to reduce the number of expert model evaluations at test time.

This paper presents a vehicle-independent, non-intrusive, and light monitoring system for accurately measuring fuel consumption in road vehicles from longitudinal speed and acceleration derived continuously in time from GNSS and IMU sensors mounted inside the vehicle. In parallel to boosting the transition to zero-carbon cars, there is an increasing interest in low-cost instruments for precise measurement of the environmental impact of the many internal combustion engine vehicles still in circulation. The main contribution of this work is the design and comparison of two innovative black-box algorithms, one based on a reduced complexity physics modeling while the other relying on a feedforward neural network for black-box fuel consumption estimation using only velocity and acceleration measurements. Based on suitable metrics, the developed algorithms outperform the state of the art best approach, both in the instantaneous and in the integral fuel consumption estimation, with errors smaller than 1\% with respect to the fuel flow ground truth. The data used for model identification, testing, and experimental validation is composed of GNSS velocity and IMU acceleration measurements collected during several trips using a diesel fuel vehicle on different roads, in different seasons, and with varying numbers of passengers. Compared to built-in vehicle monitoring systems, this methodology is not customized, uses off-the-shelf sensors, and is based on two simple algorithms that have been validated offline and could be easily implemented in a real-time environment.

An important challenge for autonomous agents such as robots is to maintain a spatially and temporally consistent model of the world. It must be maintained through occlusions, previously-unseen views, and long time horizons (e.g., loop closure and re-identification). It is still an open question how to train such a versatile neural representation without supervision. We start from the idea that the training objective can be framed as a patch retrieval problem: given an image patch in one view of a scene, we would like to retrieve (with high precision and recall) all patches in other views that map to the same real-world location. One drawback is that this objective does not promote reusability of features: by being unique to a scene (achieving perfect precision/recall), a representation will not be useful in the context of other scenes. We find that it is possible to balance retrieval and reusability by constructing the retrieval set carefully, leaving out patches that map to far-away locations. Similarly, we can easily regulate the scale of the learned features (e.g., points, objects, or rooms) by adjusting the spatial tolerance for considering a retrieval to be positive. We optimize for (smooth) Average Precision (AP), in a single unified ranking-based objective. This objective also doubles as a criterion for choosing landmarks or keypoints, as patches with high AP. We show results creating sparse, multi-scale, semantic spatial maps composed of highly identifiable landmarks, with applications in landmark retrieval, localization, semantic segmentation and instance segmentation.

Fifth-generation (5G) mobile communication networks have recently emerged in various fields, including highspeed trains. However, the dense deployment of 5G millimeter wave (mmWave) base stations (BSs) and the high speed of moving trains lead to frequent handovers (HOs), which can adversely affect the Quality-of-Service (QoS) of mobile users. As a result, HO optimization and resource allocation are essential considerations for managing mobility in high-speed train systems. In this paper, we model system performance of a high-speed train system with a novel machine learning (ML) approach that is nested cross validation scheme that prevents information leakage from model evaluation into the model parameter tuning, thereby avoiding overfitting and resulting in better generalization error. To this end, we employ ML methods for the high-speed train system scenario. Handover Margin (HOM) and Time-to-Trigger (TTT) values are used as features, and several KPIs are used as outputs, and several ML methods including Gradient Boosting Regression (GBR), Adaptive Boosting (AdaBoost), CatBoost Regression (CBR), Artificial Neural Network (ANN), Kernel Ridge Regression (KRR), Support Vector Regression (SVR), and k-Nearest Neighbor Regression (KNNR) are employed for the problem. Finally, performance comparisons of the cross validation schemes with the methods are made in terms of mean absolute error (MAE) and mean square error (MSE) metrics are made. As per obtained results, boosting methods, ABR, CBR, GBR, with nested cross validation scheme superiorly outperforms conventional cross validation scheme results with the same methods. On the other hand, SVR, KNRR, KRR, ANN with the nested scheme produce promising results for prediction of some KPIs with respect to their conventional scheme employment.