Landmines remain a threat to war-affected communities for years after conflicts have ended, partly due to the laborious nature of demining tasks. Humanitarian demining operations begin by collecting relevant information from the sites to be cleared, which is then analyzed by human experts to determine the potential risk of remaining landmines. In this paper, we propose RELand system to support these tasks, which consists of three major components. We (1) provide general feature engineering and label assigning guidelines to enhance datasets for landmine risk modeling, which are widely applicable to global demining routines, (2) formulate landmine presence as a classification problem and design a novel interpretable model based on sparse feature masking and invariant risk minimization, and run extensive evaluation under proper protocols that resemble real-world demining operations to show a significant improvement over the state-of-the-art, and (3) build an interactive web interface to suggest priority areas for demining organizations. We are currently collaborating with a humanitarian demining NGO in Colombia that is using our system as part of their field operations in two areas recently prioritized for demining.

Tactile sensing has recently been used in robotics for object identification, grasping, and material recognition. Most material recognition approaches use vibration information from a tactile exploration, typically above one second long, to identify the material. This work proposes a tactile multi-modal (vibration and thermal) material identification approach based on recursive Bayesian estimation. Through the frequency response of the vibration induced by the material and thermal features, like an estimate of the thermal power loss of the finger, we show that it is possible to identify materials in less than half a second. Moreover, a comparison between the use of vibration only and multi-modal identification shows that both recognition time and classification errors are reduced by adding thermal information.

Population-based structural health monitoring (PBSHM), seeks to address some of the limitations associated with data scarcity that arise in traditional SHM. A tenet of the population-based approach to SHM is that information can be shared between sufficiently-similar structures in order to improve predictive models. Transfer learning techniques, such as domain adaptation, have been shown to be a highly-useful technology for sharing information between structures when developing statistical classifiers for PBSHM. Nonetheless, transfer-learning techniques are not without their pitfalls. In some circumstances, for example if the data distributions associated with the structures within a population are dissimilar, applying transfer-learning methods can be detrimental to classification performance -- this phenomenon is known as negative transfer. Given the potentially-severe consequences of negative transfer, it is prudent for engineers to ask the question `when, what, and how should one transfer between structures?'. The current paper aims to demonstrate a transfer-strategy decision process for a classification task for a population of simulated structures in the context of a representative SHM maintenance problem, supported by domain adaptation. The transfer decision framework is based upon the concept of expected value of information transfer. In order to compute the expected value of information transfer, predictions must be made regarding the classification (and decision performance) in the target domain following information transfer. In order to forecast the outcome of transfers, a probabilistic regression is used here to predict classification performance from a proxy for structural similarity based on the modal assurance criterion.

Quantifying the impact of individual data samples on machine learning models is an open research problem. This is particularly relevant when complex and high-dimensional relationships have to be learned from a limited sample of the data generating distribution, such as in deep learning. It was previously shown that, in these cases, models rely not only on extracting patterns which are helpful for generalisation, but also seem to be required to incorporate some of the training data more or less as is, in a process often termed memorisation. This raises the question: if some memorisation is a requirement for effective learning, what are its privacy implications? In this work we unify a broad range of previous definitions and perspectives on memorisation in ML, discuss their interplay with model generalisation and their implications of these phenomena on data privacy. Moreover, we systematise methods allowing practitioners to detect the occurrence of memorisation or quantify it and contextualise our findings in a broad range of ML learning settings. Finally, we discuss memorisation in the context of privacy attacks, differential privacy (DP) and adversarial actors.

In recent years, process mining emerged as a proven technology to analyze and improve operational processes. An expanding range of organizations using process mining in their daily operation brings a broader spectrum of processes to be analyzed. Some of these processes are highly unstructured, making it difficult for traditional process discovery approaches to discover a start-to-end model describing the entire process. Therefore, the subdiscipline of Local Process Model (LPM) discovery tries to build a set of LPMs, i.e., smaller models that explain sub-behaviors of the process. However, like other pattern mining approaches, LPM discovery algorithms also face the problems of model explosion and model repetition, i.e., the algorithms may create hundreds if not thousands of models, and subsets of them are close in structure or behavior. This work proposes a three-step pipeline for grouping similar LPMs using various process model similarity measures. We demonstrate the usefulness of grouping through a real-life case study, and analyze the impact of different measures, the gravity of repetition in the discovered LPMs, and how it improves after grouping on multiple real event logs.

This article offers a comprehensive treatment of polynomial functional regression, culminating in the establishment of a novel finite sample bound. This bound encompasses various aspects, including general smoothness conditions, capacity conditions, and regularization techniques. In doing so, it extends and generalizes several findings from the context of linear functional regression as well. We also provide numerical evidence that using higher order polynomial terms can lead to an improved performance.

We introduce TabRepo, a new dataset of tabular model evaluations and predictions. TabRepo contains the predictions and metrics of 1206 models evaluated on 200 classification and regression datasets. We illustrate the benefit of our dataset in multiple ways. First, we show that it allows to perform analysis such as comparing Hyperparameter Optimization against current AutoML systems while also considering ensembling at no cost by using precomputed model predictions. Second, we show that our dataset can be readily leveraged to perform transfer-learning. In particular, we show that applying standard transfer-learning techniques allows to outperform current state-of-the-art tabular systems in accuracy, runtime and latency. Machine learning on structured tabular data has a long history due to its wide range of practical applications. Significant progress has been achieved through improving supervised learning models, with key method landmarks including SVM (Hearst et al., 1998), Random Forest (Breiman, 2001) and Gradient Boosted Trees (Friedman, 2001). While the performance of base models is still being improved by a steady stream of research, their performance has saturated and state-of-the-art methods now leverage AutoML techniques (He et al., 2021) or new paradigms such as the pretraining of transformer models (Hollmann et al., 2022). AutoML solutions currently dominate tabular prediction benchmarks (Erickson et al., 2020; Gijsbers et al., 2022). Auto-Sklearn (Feurer et al., 2015a; 2020) was an early approach that proposed to select pipelines to ensemble from the Sklearn library and meta-learn the hyperparameter-optimization (HPO) with offline evaluations.

We present HUME, a simple model-agnostic framework for inferring human labeling of a given dataset without any external supervision. The key insight behind our approach is that classes defined by many human labelings are linearly separable regardless of the representation space used to represent a dataset. HUME utilizes this insight to guide the search over all possible labelings of a dataset to discover an underlying human labeling. We show that the proposed optimization objective is strikingly well-correlated with the ground truth labeling of the dataset. In effect, we only train linear classifiers on top of pretrained representations that remain fixed during training, making our framework compatible with any large pretrained and self-supervised model. Despite its simplicity, HUME outperforms a supervised linear classifier on top of self-supervised representations on the STL-10 dataset by a large margin and achieves comparable performance on the CIFAR-10 dataset. Compared to the existing unsupervised baselines, HUME achieves state-of-the-art performance on four benchmark image classification datasets including the large-scale ImageNet-1000 dataset. Altogether, our work provides a fundamentally new view to tackle unsupervised learning by searching for consistent labelings between different representation spaces.

Localisation in GPS-denied environments is challenging and many existing solutions have infrastructural and on-site calibration requirements. This paper tackles these challenges by proposing a localisation system that is infrastructure-free and does not require on-site calibration, using a single active PTZ camera to detect, track and localise a circular LED marker. We propose to use a CNN trained using only synthetic images to detect the LED marker as an ellipse and show that our approach is more robust than using traditional ellipse detection without requiring tuning of parameters for feature extraction. We also propose to leverage the predicted elliptical angle as a measure of uncertainty of the CNN's predictions and show how it can be used in a filter to improve marker range estimation and 3D localisation. We evaluate our system's performance through localisation of a UAV in real-world flight experiments and show that it can outperform alternative methods for localisation in GPS-denied environments. We also demonstrate our system's performance in indoor and outdoor environments.

The human brain can be considered as complex networks, composed of various regions that continuously exchange their information with each other, forming the brain network graph, from which nodes and edges are extracted using resting-state functional magnetic resonance imaging (rs-fMRI). Therefore, this graph can potentially depict abnormal patterns that have emerged under the influence of brain disorders. So far, numerous studies have attempted to find embeddings for brain network graphs and subsequently classify samples with brain disorders from healthy ones, which include limitations such as: not considering the relationship between samples, not utilizing phenotype information, lack of temporal analysis, using static functional connectivity (FC) instead of dynamic ones and using a fixed graph structure. We propose a hierarchical dynamic graph representation learning (HDGL) model, which is the first model designed to address all the aforementioned challenges. HDGL consists of two levels, where at the first level, it constructs brain network graphs and learns their spatial and temporal embeddings, and at the second level, it forms population graphs and performs classification after embedding learning. Furthermore, based on how these two levels are trained, four methods have been introduced, some of which are suggested for reducing memory complexity. We evaluated the performance of the proposed model on the ABIDE and ADHD-200 datasets, and the results indicate the improvement of this model compared to several state-of-the-art models in terms of various evaluation metrics.