我们表明，如果考虑密度感知的认知不确定性项，则有效地量化神经辐射场（NERF）中的模型不确定性。在先前的工作中调查的幼稚合奏简单地渲染了RGB图像，以量化因观察到的场景的解释而引起的模型不确定性。相比之下，我们还考虑了各个射线沿线的终止概率，以确定认知模型的不确定性，因为对训练过程中未观察到的场景部分的知识不足。我们在NERF的既定不确定性量化基准中实现了新的最先进的性能，优于需要对NERF架构和培训制度进行复杂更改的方法。我们此外表明，可以将NERF不确定性用于次要视图选择和模型改进。

在对象检测中，当检测器未能检测到目标对象时，会出现假阴性。为了了解为什么对象检测产生假阴性，我们确定了五个“假负机制”，其中每个机制都描述了检测器体系结构内部的特定组件如何失败。着眼于两阶段和一阶段锚点对象检测器体系结构，我们引入了一个框架，用于量化这些虚假的负面机制。使用此框架，我们调查了为什么更快的R-CNN和视网膜无法检测基准视觉数据集和机器人数据集中的对象。我们表明，检测器的假负机制在计算机视觉基准数据集和机器人部署方案之间存在显着差异。这对为机器人应用程序开发的对象检测器的翻译具有影响。

部署到开放世界中，对象探测器容易出现开放式错误，训练数据集中不存在的对象类的假阳性检测。我们提出了GMM-DET，一种用于从对象探测器中提取认知不确定性的实时方法，以识别和拒绝开放式错误。 GMM-DID列达探测器以产生与特定于类高斯混合模型建模的结构化的Logit空间。在测试时间时，通过所有高斯混合模型下的低对数概率识别开放式错误。我们测试了两个常见的探测器架构，更快的R-CNN和RETINANET，跨越了三种不同的数据集，跨越机器人和计算机视觉。我们的结果表明，GMM-DET始终如一地优于识别和拒绝开放式检测的现有不确定性技术，特别是在安全关键应用程序所需的低差错率操作点。 GMM-DET保持对象检测性能，并仅引入最小的计算开销。我们还介绍一种用于将现有对象检测数据集转换为特定的开放式数据集的方法，以评估对象检测中的开放式性能。

Regularising the parameter matrices of neural networks is ubiquitous in training deep models. Typical regularisation approaches suggest initialising weights using small random values, and to penalise weights to promote sparsity. However, these widely used techniques may be less effective in certain scenarios. Here, we study the Koopman autoencoder model which includes an encoder, a Koopman operator layer, and a decoder. These models have been designed and dedicated to tackle physics-related problems with interpretable dynamics and an ability to incorporate physics-related constraints. However, the majority of existing work employs standard regularisation practices. In our work, we take a step toward augmenting Koopman autoencoders with initialisation and penalty schemes tailored for physics-related settings. Specifically, we propose the "eigeninit" initialisation scheme that samples initial Koopman operators from specific eigenvalue distributions. In addition, we suggest the "eigenloss" penalty scheme that penalises the eigenvalues of the Koopman operator during training. We demonstrate the utility of these schemes on two synthetic data sets: a driven pendulum and flow past a cylinder; and two real-world problems: ocean surface temperatures and cyclone wind fields. We find on these datasets that eigenloss and eigeninit improves the convergence rate by up to a factor of 5, and that they reduce the cumulative long-term prediction error by up to a factor of 3. Such a finding points to the utility of incorporating similar schemes as an inductive bias in other physics-related deep learning approaches.

The number of international benchmarking competitions is steadily increasing in various fields of machine learning (ML) research and practice. So far, however, little is known about the common practice as well as bottlenecks faced by the community in tackling the research questions posed. To shed light on the status quo of algorithm development in the specific field of biomedical imaging analysis, we designed an international survey that was issued to all participants of challenges conducted in conjunction with the IEEE ISBI 2021 and MICCAI 2021 conferences (80 competitions in total). The survey covered participants' expertise and working environments, their chosen strategies, as well as algorithm characteristics. A median of 72% challenge participants took part in the survey. According to our results, knowledge exchange was the primary incentive (70%) for participation, while the reception of prize money played only a minor role (16%). While a median of 80 working hours was spent on method development, a large portion of participants stated that they did not have enough time for method development (32%). 25% perceived the infrastructure to be a bottleneck. Overall, 94% of all solutions were deep learning-based. Of these, 84% were based on standard architectures. 43% of the respondents reported that the data samples (e.g., images) were too large to be processed at once. This was most commonly addressed by patch-based training (69%), downsampling (37%), and solving 3D analysis tasks as a series of 2D tasks. K-fold cross-validation on the training set was performed by only 37% of the participants and only 50% of the participants performed ensembling based on multiple identical models (61%) or heterogeneous models (39%). 48% of the respondents applied postprocessing steps.

Adversarial attacks hamper the decision-making ability of neural networks by perturbing the input signal. The addition of calculated small distortion to images, for instance, can deceive a well-trained image classification network. In this work, we propose a novel attack technique called Sparse Adversarial and Interpretable Attack Framework (SAIF). Specifically, we design imperceptible attacks that contain low-magnitude perturbations at a small number of pixels and leverage these sparse attacks to reveal the vulnerability of classifiers. We use the Frank-Wolfe (conditional gradient) algorithm to simultaneously optimize the attack perturbations for bounded magnitude and sparsity with $O(1/\sqrt{T})$ convergence. Empirical results show that SAIF computes highly imperceptible and interpretable adversarial examples, and outperforms state-of-the-art sparse attack methods on the ImageNet dataset.

Although many machine learning methods, especially from the field of deep learning, have been instrumental in addressing challenges within robotic applications, we cannot take full advantage of such methods before these can provide performance and safety guarantees. The lack of trust that impedes the use of these methods mainly stems from a lack of human understanding of what exactly machine learning models have learned, and how robust their behaviour is. This is the problem the field of explainable artificial intelligence aims to solve. Based on insights from the social sciences, we know that humans prefer contrastive explanations, i.e.\ explanations answering the hypothetical question "what if?". In this paper, we show that linear model trees are capable of producing answers to such questions, so-called counterfactual explanations, for robotic systems, including in the case of multiple, continuous inputs and outputs. We demonstrate the use of this method to produce counterfactual explanations for two robotic applications. Additionally, we explore the issue of infeasibility, which is of particular interest in systems governed by the laws of physics.

Decentralized learning with private data is a central problem in machine learning. We propose a novel distillation-based decentralized learning technique that allows multiple agents with private non-iid data to learn from each other, without having to share their data, weights or weight updates. Our approach is communication efficient, utilizes an unlabeled public dataset and uses multiple auxiliary heads for each client, greatly improving training efficiency in the case of heterogeneous data. This approach allows individual models to preserve and enhance performance on their private tasks while also dramatically improving their performance on the global aggregated data distribution. We study the effects of data and model architecture heterogeneity and the impact of the underlying communication graph topology on learning efficiency and show that our agents can significantly improve their performance compared to learning in isolation.

Many current approaches to machine learning in particle physics use generic architectures that require large numbers of parameters and disregard underlying physics principles, limiting their applicability as scientific modeling tools. In this work, we present a machine learning architecture that uses a set of inputs maximally reduced with respect to the full 6-dimensional Lorentz symmetry, and is fully permutation-equivariant throughout. We study the application of this network architecture to the standard task of top quark tagging and show that the resulting network outperforms all existing competitors despite much lower model complexity. In addition, we present a Lorentz-covariant variant of the same network applied to a 4-momentum regression task.

This paper investigates how hate speech varies in systematic ways according to the identities it targets. Across multiple hate speech datasets annotated for targeted identities, we find that classifiers trained on hate speech targeting specific identity groups struggle to generalize to other targeted identities. This provides empirical evidence for differences in hate speech by target identity; we then investigate which patterns structure this variation. We find that the targeted demographic category (e.g. gender/sexuality or race/ethnicity) appears to have a greater effect on the language of hate speech than does the relative social power of the targeted identity group. We also find that words associated with hate speech targeting specific identities often relate to stereotypes, histories of oppression, current social movements, and other social contexts specific to identities. These experiments suggest the importance of considering targeted identity, as well as the social contexts associated with these identities, in automated hate speech classification.