To reveal the mechanical behavior and deformation patterns of geotechnical reinforcement materials under tensile loading, a series tests were conducted on plastic geogrid rib, fiberglass gabion steel wire, mesh, mesh. The full force-strain relationships obtained. failure modes different discussed. standard linear three-element model, nonlinear improved Kawabata model employed to simulate curves various materials. main parameters models determined. results showed that brittle occurred in both...

In this work, we develop a novel fairness learning approach for multi-task regression models based on biased training dataset, using popular rank-based non-parametric independence test, i.e., Mann Whitney U statistic, measuring the dependency between target variable and protected variables. To solve problem efficiently, first reformulate as new non-convex optimization problem, in which constraint is defined group-wise ranking functions of individual objects. We then an efficient...

In contrast to offline working fashions, two research paradigms are devised for online learning: (1) Online Meta-Learning (OML)[6, 20, 26] learns good priors over model parameters (or learning learn) in a sequential setting where tasks revealed one after another. Although it provides sub-linear regret bound, such techniques completely ignore the importance of with fairness which is significant hallmark human intelligence. (2) Fairness-Aware Learning [1, 8, 21]. This captures many...

Self-supervised learning has recently achieved great success in representation without human annotations. The dominant method – that is contrastive learning, generally based on instance discrimination tasks, i.e., individual samples are treated as independent categories. However, presuming all the different contradicts natural grouping of similar common visual datasets, e.g., multiple views same dog. To bridge gap, this paper proposes an adaptive introduces soft inter-sample relations,...

Scene completion and forecasting are two popular perception problems in research for mobile agents like autonomous vehicles. Existing approaches treat the isolation, resulting a separate of aspects. In this paper, we introduce novel LiDAR task Occupancy Completion Forecasting (OCF) context driving to unify these aspects into cohesive framework. This requires new algorithms address three challenges altogether: (1) sparse-to-dense reconstruction, (2) partial-to-complete hallucination, (3)...

Learning disentangled causal representations is a challenging problem that has gained significant attention recently due to its implications for extracting meaningful information downstream tasks. In this work, we define new notion of disentanglement from the perspective independent mechanisms. We propose ICM-VAE, framework learning causally supervised by related observed labels. model mechanisms using nonlinear learnable flow-based diffeomorphic functions map noise variables latent...

Recent multilingual pre-trained language models have achieved remarkable zero-shot performance, where the model is only finetuned on one source and directly evaluated target languages. In this work, we propose a self-learning framework that further utilizes unlabeled data of languages, combined with uncertainty estimation in process to select high-quality silver labels. Three different uncertainties are adapted analyzed specifically for cross lingual transfer: Language...

Pretrained transformers exhibit the remarkable ability of in-context learning (ICL): they can learn tasks from just a few examples provided in prompt without updating any weights. This raises foundational question: ICL solve fundamentally $\textit{new}$ that are very different those seen during pretraining? To probe this question, we examine ICL's performance on linear regression while varying diversity pretraining dataset. We empirically demonstrate $\textit{task threshold}$ for emergence...

Many C programs assume the use of implicit domain-specific information. A common example is units measurement, where values can have both a standard type and an associated unit. However, since there no way in language to represent this additional information, violations policies, such as unit safety violations, be difficult detect. In paper we present static analysis, based on abstract semantics defined using rewriting logic, for detection programs. contrast typed approaches, analysis makes...

Artificial intelligence nowadays plays an increasingly prominent role in our life since decisions that were once made by humans are now delegated to automated systems. A machine learning algorithm trained based on biased data, however, tends make unfair predictions. Developing classification algorithms fair with respect protected attributes of the data thus becomes important problem. Motivated concerns surrounding fairness effects sharing and few-shot tools, such as Model Agnostic...

The problem of learning to generalize on unseen classes during the training step, also known as few-shot classification, has attracted considerable attention. Initialization based methods, such gradient-based model agnostic meta-learning (MAML) [1], tackle by "learning fine-tune". goal these approaches is learn proper initialization, so that classifiers for new can be learned from a few labeled examples with small number gradient update steps. Few shot well-known its fast-adapted capability...

Domain generalization (DG) aims to improve the ability of model trained on several known training domains over unseen test domains. Previous work has shown that self-supervised contrastive pre-training improves robustness downstream tasks. However, in this paper, we find models do not exhibit better performance than supervised pre-trained same dataset DG setting. We argue is owing fact richer intra-class discriminative features extracted by learning, which term silent features, are...

Despite the large progress in supervised learning with neural networks, there are significant challenges obtaining high-quality, large-scale and accurately labelled datasets. In such a context, how to learn presence of noisy labels has received more attention. As relatively complex problem, order achieve good results, current approaches often integrate components from several fields, as learning, semi-supervised transfer resulting complicated methods. Furthermore, they make multiple...

Based on the development of lighting control and management system in central area Beijing Olympic park, a large-scale scheme design language (LALSDL) has been proposed. Through analysis demand manage devices characteristics that LALSDL should have are summarized; through summary involved operations device declaration operation statement defined; call expression (DCE) is used to define procedure device, The operational semantics DCE strictly defined. results this paper applied which was good...

The fairness-aware online learning framework has emerged as a potent tool within the context of continuous lifelong learning. In this scenario, learner's objective is to progressively acquire new tasks they arrive over time, while also guaranteeing statistical parity among various protected sub-populations, such race and gender, when it comes newly introduced tasks. A significant limitation current approaches lies in their heavy reliance on i.i.d (independent identically distributed)...

<abstract><p>With the widespread adoption of electronic health records, amount stored medical data has been increasing. Clinical data, often in form semi-structured or unstructured records (EMRs), contains rich patient information. However, due to use natural language by physicians when composing these effectiveness traditional methods such as dictionaries, rule matching, and machine learning extraction information from texts falls short clinical standards. In this paper, a novel...

Learning with Noisy labels (LNL) poses a significant challenge for the Machine community. Some of most widely used approaches that select as clean samples which model itself (the in-training model) has high confidence, e.g., `small loss', can suffer from so called `self-confirmation' bias. This bias arises because model, is at least partially trained on noisy labels. Furthermore, in classification case, an additional some label noise between classes are visually very similar (`hard noise')....

Generalizing to out-of-distribution data while being aware of model fairness is a significant and challenging problem in meta-learning. The goal this find set fairness-aware invariant parameters classifier that trained using drawn from family related training domains with distribution shift on non-sensitive features as well different levels dependence between predictions sensitive so the can achieve good generalization performance unknown but distinct test domains. To tackle challenge,...

Given the inherent non-stationarity prevalent in real-world applications, continual Reinforcement Learning (RL) aims to equip agent with capability address a series of sequentially presented decision-making tasks. Within this problem setting, pivotal challenge revolves around \textit{catastrophic forgetting} issue, wherein is prone effortlessly erode decisional knowledge associated past encountered tasks when learning new one. In recent progresses, \textit{generative replay} methods have...