Neural networks are part of daily-life decision-making, including in high-stakes settings where understanding and transparency are key. Saliency maps have been developed to gain understanding into which input features neural networks use for a specific prediction. Although widely employed, these methods often result in overly general saliency maps that fail to identify the specific information that triggered the classification. In this work, we suggest a framework that allows to incorporate attributions across classes to arrive at saliency maps that actually capture the class-relevant information. On established benchmarks for attribution methods, including the grid-pointing game and randomization-based sanity checks, we show that our framework heavily boosts the performance of standard saliency map approaches. It is, by design, agnostic to model architectures and attribution methods and now allows to identify the distinguishing and shared features used for a model prediction.

Machine learning models deployed in sensitive areas such as healthcare must be interpretable to ensure accountability and fairness. However, learning such rule lists presents significant challenges. Existing methods based on combinatorial optimization require feature pre-discretization and impose restrictions on rule size. Neuro-symbolic methods use more scalable continuous optimization yet place similar pre-discretization constraints and suffer from unstable optimization. We formulate a continuous relaxation of the rule list learning problem that converges to a strict rule list through temperature annealing. Machine learning models are increasingly used in high-stakes applications such as healthcare (Deo, 2015), credit risk evaluation (Bhatore et al., 2020), and criminal justice (Lakkaraju & Rudin, 2017), where it is vital that each decision is fair and reasonable. Proxy measures such as Shapley values can give the illusion of interpretability, but are highly problematic as they can not faithfully represent a non-additive models decision process (Gosiewska & Biecek, 2019). Instead, Rudin (2019) argues that it is crucial to use inherently interpretable models, to create systems with human supervision in the loop (Kleinberg et al., 2018). For particularly sensitive domains such as stroke prediction or recidivism, so called Rule Lists are a popular choice (Letham et al., 2015) due to their fully transparent decision making. A rule list predicts based on nested "if-then-else" statements and naturally aligns with the human-decision making process. Each rule is active if its conditions are met, e.g. " if Thalassemia = normal Resting bps < 151 ", and carries a respective prediction, i.e. " then P ( Disease) = 10% ".

Flatness of the loss surface not only correlates positively with generalization but is also related to adversarial robustness, since perturbations of inputs relate non-linearly to perturbations of weights. In this paper, we empirically analyze the relation between adversarial examples and relative flatness with respect to the parameters of one layer. We observe a peculiar property of adversarial examples: during an iterative first-order white-box attack, the flatness of the loss surface measured around the adversarial example first becomes sharper until the label is flipped, but if we keep the attack running it runs into a flat uncanny valley where the label remains flipped. We find this phenomenon across various model architectures and datasets. Our results also extend to large language models (LLMs), but due to the discrete nature of the input space and comparatively weak attacks, the adversarial examples rarely reach a truly flat region. Most importantly, this phenomenon shows that flatness alone cannot explain adversarial robustness unless we can also guarantee the behavior of the function around the examples. We theoretically connect relative flatness to adversarial robustness by bounding the third derivative of the loss surface, underlining the need for flatness in combination with a low global Lipschitz constant for a robust model.

Finding and describing sub-populations that are exceptional regarding a target property has important applications in many scientific disciplines, from identifying disadvantaged demographic groups in census data to finding conductive molecules within gold nanoparticles. Current approaches to finding such subgroups require pre-discretized predictive variables, do not permit non-trivial target distributions, do not scale to large datasets, and struggle to find diverse results. To address these limitations, we propose Syflow, an end-to-end optimizable approach in which we leverage normalizing flows to model arbitrary target distributions, and introduce a novel neural layer that results in easily interpretable subgroup descriptions. We demonstrate on synthetic and real-world data, including a case study, that Syflow reliably finds highly exceptional subgroups accompanied by insightful descriptions.

Discovering patterns in data that best describe the differences between classes allows to hypothesize and reason about class-specific mechanisms. In molecular biology, for example, this bears promise of advancing the understanding of cellular processes differing between tissues or diseases, which could lead to novel treatments. To be useful in practice, methods that tackle the problem of finding such differential patterns have to be readily interpretable by domain experts, and scalable to the extremely high-dimensional data. In this work, we propose a novel, inherently interpretable binary neural network architecture DIFFNAPS that extracts differential patterns from data. DiffNaps is scalable to hundreds of thousands of features and robust to noise, thus overcoming the limitations of current state-of-the-art methods in large-scale applications such as in biology. We show on synthetic and real world data, including three biological applications, that, unlike its competitors, DiffNaps consistently yields accurate, succinct, and interpretable class descriptions

Modern deep learning architecture utilize batch normalization (BN) to stabilize training and improve accuracy. It has been shown that the BN layers alone are surprisingly expressive. In the context of robustness against adversarial examples, however, BN is argued to increase vulnerability. That is, BN helps to learn fragile features. Nevertheless, BN is still used in adversarial training, which is the de-facto standard to learn robust features. In order to shed light on the role of BN in adversarial training, we investigate to what extent the expressiveness of BN can be used to robustify fragile features in comparison to random features. On CIFAR10, we find that adversarially fine-tuning just the BN layers can result in non-trivial adversarial robustness. Adversarially training only the BN layers from scratch, in contrast, is not able to convey meaningful adversarial robustness. Our results indicate that fragile features can be used to learn models with moderate adversarial robustness, while random features cannot