Search

Chronic rhinosinusitis (CRS) is a common and persistent sinus imflammation that affects 5 - 12\% of the general population. It significantly impacts quality of life and is often difficult to assess due to its subjective nature in clinical evaluation. We introduce PARASIDE, an automatic tool for segmenting air and soft tissue volumes of the structures of the sinus maxillaris, frontalis, sphenodalis and ethmoidalis in T1 MRI. By utilizing that segmentation, we can quantify feature relations that have been observed only manually and subjectively before. We performed an exemplary study and showed both volume and intensity relations between structures and radiology reports. While the soft tissue segmentation is good, the automated annotations of the air volumes are excellent. The average intensity over air structures are consistently below those of the soft tissues, close to perfect separability. Healthy subjects exhibit lower soft tissue volumes and lower intensities. Our developed system is the first automated whole nasal segmentation of 16 structures, and capable of calculating medical relevant features such as the Lund-Mackay score.

Machine learning-based classifiers are commonly evaluated by metrics like accuracy, but deeper analysis is required to understand their strengths and weaknesses. MLMC is a visual exploration tool that tackles the challenge of multi-label classifier comparison and evaluation. It offers a scalable alternative to confusion matrices which are commonly used for such tasks, but don't scale well with a large number of classes or labels. Additionally, MLMC allows users to view classifier performance from an instance perspective, a label perspective, and a classifier perspective. Our user study shows that the techniques implemented by MLMC allow for a powerful multi-label classifier evaluation while preserving user friendliness., Comment: 12 pages

Benchmarks are important tools for tracking the rapid advancements in large language model (LLM) capabilities. However, benchmarks are not keeping pace in difficulty: LLMs now achieve over 90\% accuracy on popular benchmarks like MMLU, limiting informed measurement of state-of-the-art LLM capabilities. In response, we introduce Humanity's Last Exam (HLE), a multi-modal benchmark at the frontier of human knowledge, designed to be the final closed-ended academic benchmark of its kind with broad subject coverage. HLE consists of 3,000 questions across dozens of subjects, including mathematics, humanities, and the natural sciences. HLE is developed globally by subject-matter experts and consists of multiple-choice and short-answer questions suitable for automated grading. Each question has a known solution that is unambiguous and easily verifiable, but cannot be quickly answered via internet retrieval. State-of-the-art LLMs demonstrate low accuracy and calibration on HLE, highlighting a significant gap between current LLM capabilities and the expert human frontier on closed-ended academic questions. To inform research and policymaking upon a clear understanding of model capabilities, we publicly release HLE at https://lastexam.ai., Comment: 25 pages, 6 figures

Counterfactual examples are widely used in natural language processing (NLP) as valuable data to improve models, and in explainable artificial intelligence (XAI) to understand model behavior. The automated generation of counterfactual examples remains a challenging task even for large language models (LLMs), despite their impressive performance on many tasks. In this paper, we first introduce ZeroCF, a faithful approach for leveraging important words derived from feature attribution methods to generate counterfactual examples in a zero-shot setting. Second, we present a new framework, FitCF, which further verifies aforementioned counterfactuals by label flip verification and then inserts them as demonstrations for few-shot prompting, outperforming two state-of-the-art baselines. Through ablation studies, we identify the importance of each of FitCF's core components in improving the quality of counterfactuals, as assessed through flip rate, perplexity, and similarity measures. Furthermore, we show the effectiveness of LIME and Integrated Gradients as backbone attribution methods for FitCF and find that the number of demonstrations has the largest effect on performance. Finally, we reveal a strong correlation between the faithfulness of feature attribution scores and the quality of generated counterfactuals., Comment: In submission

Being able to reconstruct training data from the parameters of a neural network is a major privacy concern. Previous works have shown that reconstructing training data, under certain circumstances, is possible. In this work, we analyse such reconstructions empirically and propose a new formulation of the reconstruction as a solution to a bilevel optimisation problem. We demonstrate that our formulation as well as previous approaches highly depend on the initialisation of the training images $x$ to reconstruct. In particular, we show that a random initialisation of $x$ can lead to reconstructions that resemble valid training samples while not being part of the actual training dataset. Thus, our experiments on affine and one-hidden layer networks suggest that when reconstructing natural images, yet an adversary cannot identify whether reconstructed images have indeed been part of the set of training samples.

Lifted probabilistic inference exploits symmetries in a probabilistic model to allow for tractable probabilistic inference with respect to domain sizes of logical variables. We found that the current state-of-the-art algorithm to construct a lifted representation in form of a parametric factor graph misses symmetries between factors that are exchangeable but scaled differently, thereby leading to a less compact representation. In this paper, we propose a generalisation of the advanced colour passing (ACP) algorithm, which is the state of the art to construct a parametric factor graph. Our proposed algorithm allows for potentials of factors to be scaled arbitrarily and efficiently detects more symmetries than the original ACP algorithm. By detecting strictly more symmetries than ACP, our algorithm significantly reduces online query times for probabilistic inference when the resulting model is applied, which we also confirm in our experiments., Comment: Accepted to the Proceedings of the 3rd Learning on Graphs Conference (LoG 2024)

Lifting uses a representative of indistinguishable individuals to exploit symmetries in probabilistic relational models, denoted as parametric factor graphs, to speed up inference while maintaining exact answers. In this paper, we show how lifting can be applied to causal inference in partially directed graphs, i.e., graphs that contain both directed and undirected edges to represent causal relationships between random variables. We present partially directed parametric causal factor graphs (PPCFGs) as a generalisation of previously introduced parametric causal factor graphs, which require a fully directed graph. We further show how causal inference can be performed on a lifted level in PPCFGs, thereby extending the applicability of lifted causal inference to a broader range of models requiring less prior knowledge about causal relationships., Comment: Accepted to the Proceedings of the 16th International Conference on Scalable Uncertainty Management (SUM 2024)

GPT-4o is an autoregressive omni model that accepts as input any combination of text, audio, image, and video, and generates any combination of text, audio, and image outputs. It's trained end-to-end across text, vision, and audio, meaning all inputs and outputs are processed by the same neural network. GPT-4o can respond to audio inputs in as little as 232 milliseconds, with an average of 320 milliseconds, which is similar to human response time in conversation. It matches GPT-4 Turbo performance on text in English and code, with significant improvement on text in non-English languages, while also being much faster and 50\% cheaper in the API. GPT-4o is especially better at vision and audio understanding compared to existing models. In line with our commitment to building AI safely and consistent with our voluntary commitments to the White House, we are sharing the GPT-4o System Card, which includes our Preparedness Framework evaluations. In this System Card, we provide a detailed look at GPT-4o's capabilities, limitations, and safety evaluations across multiple categories, focusing on speech-to-speech while also evaluating text and image capabilities, and measures we've implemented to ensure the model is safe and aligned. We also include third-party assessments on dangerous capabilities, as well as discussion of potential societal impacts of GPT-4o's text and vision capabilities.

This work addresses the problem of \textit{shape completion}, i.e., the task of restoring incomplete shapes by predicting their missing parts. While previous works have often predicted the fractured and restored shape in one step, we approach the task by separately predicting the fractured and newly restored parts, but ensuring these predictions are interconnected. We use a decoder network motivated by related work on the prediction of signed distance functions (DeepSDF). In particular, our representation allows us to consider test-time-training, i.e., finetuning network parameters to match the given incomplete shape more accurately during inference. While previous works often have difficulties with artifacts around the fracture boundary, we demonstrate that our overfitting to the fractured parts leads to significant improvements in the restoration of eight different shape categories of the ShapeNet data set in terms of their chamfer distances.

Understanding the neurobiology of opioid use disorder (OUD) using resting-state functional magnetic resonance imaging (rs-fMRI) may help inform treatment strategies to improve patient outcomes. Recent literature suggests temporal characteristics of rs-fMRI blood oxygenation level-dependent (BOLD) signals may offer complementary information to functional connectivity analysis. However, existing studies of OUD analyze BOLD signals using measures computed across all time points. This study, for the first time in the literature, employs data-driven machine learning (ML) modeling of rs-fMRI BOLD features representing multiple time points to identify region(s) of interest that differentiate OUD subjects from healthy controls (HC). Following the triple network model, we obtain rs-fMRI BOLD features from the default mode network (DMN), salience network (SN), and executive control network (ECN) for 31 OUD and 45 HC subjects. Then, we use the Boruta ML algorithm to identify statistically significant BOLD features that differentiate OUD from HC, identifying the DMN as the most salient functional network for OUD. Furthermore, we conduct brain activity mapping, showing heightened neural activity within the DMN for OUD. We perform 5-fold cross-validation classification (OUD vs. HC) experiments to study the discriminative power of functional network features with and without fusing demographic features. The DMN shows the most discriminative power, achieving mean AUC and F1 scores of 80.91% and 73.97%, respectively, when fusing BOLD and demographic features. Follow-up Boruta analysis using BOLD features extracted from the medial prefrontal cortex, posterior cingulate cortex, and left and right temporoparietal junctions reveals significant features for all four functional hubs within the DMN., Comment: 20 pages, 7 figures, 8 tables

Natural language explanations (NLEs) are vital for elucidating the reasoning behind large language model (LLM) decisions. Many techniques have been developed to generate NLEs using LLMs. However, like humans, LLMs might not always produce optimal NLEs on first attempt. Inspired by human learning processes, we introduce Cross-Refine, which employs role modeling by deploying two LLMs as generator and critic, respectively. The generator outputs a first NLE and then refines this initial explanation using feedback and suggestions provided by the critic. Cross-Refine does not require any supervised training data or additional training. We validate Cross-Refine across three NLP tasks using three state-of-the-art open-source LLMs through automatic and human evaluation. We select Self-Refine (Madaan et al., 2023) as the baseline, which only utilizes self-feedback to refine the explanations. Our findings from automatic evaluation and a user study indicate that Cross-Refine outperforms Self-Refine. Meanwhile, Cross-Refine can perform effectively with less powerful LLMs, whereas Self-Refine only yields strong results with ChatGPT. Additionally, we conduct an ablation study to assess the importance of feedback and suggestions. Both of them play an important role in refining explanations. We further evaluate Cross-Refine on a bilingual dataset in English and German., Comment: Accepted at COLING 2025; long paper

Probabilistic relational models provide a well-established formalism to combine first-order logic and probabilistic models, thereby allowing to represent relationships between objects in a relational domain. At the same time, the field of artificial intelligence requires increasingly large amounts of relational training data for various machine learning tasks. Collecting real-world data, however, is often challenging due to privacy concerns, data protection regulations, high costs, and so on. To mitigate these challenges, the generation of synthetic data is a promising approach. In this paper, we solve the problem of generating synthetic relational data via probabilistic relational models. In particular, we propose a fully-fledged pipeline to go from relational database to probabilistic relational model, which can then be used to sample new synthetic relational data points from its underlying probability distribution. As part of our proposed pipeline, we introduce a learning algorithm to construct a probabilistic relational model from a given relational database., Comment: Accepted to the Proceedings of the 47th German Conference on Artificial Intelligence (KI 2024)

Accurate calibration of finite element (FE) models is essential across various biomechanical applications, including human intervertebral discs (IVDs), to ensure their reliability and use in diagnosing and planning treatments. However, traditional calibration methods are computationally intensive, requiring iterative, derivative-free optimization algorithms that often take days to converge. This study addresses these challenges by introducing a novel, efficient, and effective calibration method demonstrated on a human L4-L5 IVD FE model as a case study using a neural network (NN) surrogate. The NN surrogate predicts simulation outcomes with high accuracy, outperforming other machine learning models, and significantly reduces the computational cost associated with traditional FE simulations. Next, a Projected Gradient Descent (PGD) approach guided by gradients of the NN surrogate is proposed to efficiently calibrate FE models. Our method explicitly enforces feasibility with a projection step, thus maintaining material bounds throughout the optimization process. The proposed method is evaluated against SOTA Genetic Algorithm and inverse model baselines on synthetic and in vitro experimental datasets. Our approach demonstrates superior performance on synthetic data, achieving an MAE of 0.06 compared to the baselines' MAE of 0.18 and 0.54, respectively. On experimental specimens, our method outperforms the baseline in 5 out of 6 cases. While our approach requires initial dataset generation and surrogate training, these steps are performed only once, and the actual calibration takes under three seconds. In contrast, traditional calibration time scales linearly with the number of specimens, taking up to 8 days in the worst-case. Such efficiency paves the way for applying more complex FE models, potentially extending beyond IVDs, and enabling accurate patient-specific simulations., Comment: In review. Project code: https://github.com/matanat/IVD-CalibNN/

Automated Machine Learning (AutoML) significantly simplifies the deployment of machine learning models by automating tasks from data preprocessing to model selection to ensembling. AutoML systems for tabular data often employ post hoc ensembling, where multiple models are combined to improve predictive accuracy. This typically results in longer inference times, a major limitation in practical deployments. Addressing this, we introduce a hardware-aware ensemble selection approach that integrates inference time into post hoc ensembling. By leveraging an existing framework for ensemble selection with quality diversity optimization, our method evaluates ensemble candidates for their predictive accuracy and hardware efficiency. This dual focus allows for a balanced consideration of accuracy and operational efficiency. Thus, our approach enables practitioners to choose from a Pareto front of accurate and efficient ensembles. Our evaluation using 83 classification datasets shows that our approach sustains competitive accuracy and can significantly improve ensembles' operational efficiency. The results of this study provide a foundation for extending these principles to additional hardware constraints, setting the stage for the development of more resource-efficient AutoML systems., Comment: Accepted at Third International Conference on Automated Machine Learning (AutoML 2024), Workshop Track; for code, see https://github.com/Atraxus/HA-ES

Counterfactual explanations (CEs) aim to enhance the interpretability of machine learning models by illustrating how alterations in input features would affect the resulting predictions. Common CE approaches require an additional model and are typically constrained to binary counterfactuals. In contrast, we propose a novel method that operates directly on the latent space of a generative model, specifically a Diffusion Autoencoder (DAE). This approach offers inherent interpretability by enabling the generation of CEs and the continuous visualization of the model's internal representation across decision boundaries. Our method leverages the DAE's ability to encode images into a semantically rich latent space in an unsupervised manner, eliminating the need for labeled data or separate feature extraction models. We show that these latent representations are helpful for medical condition classification and the ordinal regression of severity pathologies, such as vertebral compression fractures (VCF) and diabetic retinopathy (DR). Beyond binary CEs, our method supports the visualization of ordinal CEs using a linear model, providing deeper insights into the model's decision-making process and enhancing interpretability. Experiments across various medical imaging datasets demonstrate the method's advantages in interpretability and versatility. The linear manifold of the DAE's latent space allows for meaningful interpolation and manipulation, making it a powerful tool for exploring medical image properties. Our code is available at https://doi.org/10.5281/zenodo.13859266., Comment: Accepted for publication at the Journal of Machine Learning for Biomedical Imaging (MELBA) https://melba-journal.org/2024:024. arXiv admin note: text overlap with arXiv:2303.12031

Collaborative causal inference (CCI) is a federated learning method for pooling data from multiple, often self-interested, parties, to achieve a common learning goal over causal structures, e.g. estimation and optimization of treatment variables in a medical setting. Since obtaining data can be costly for the participants and sharing unique data poses the risk of losing competitive advantages, motivating the participation of all parties through equitable rewards and incentives is necessary. This paper devises an evaluation scheme to measure the value of each party's data contribution to the common learning task, tailored to causal inference's statistical demands, by comparing completed partially directed acyclic graphs (CPDAGs) inferred from observational data contributed by the participants. The Data Valuation Scheme thus obtained can then be used to introduce mechanisms that incentivize the agents to contribute data. It can be leveraged to reward agents fairly, according to the quality of their data, or to maximize all agents' data contributions.

An adverse drug effect (ADE) is any harmful event resulting from medical drug treatment. Despite their importance, ADEs are often under-reported in official channels. Some research has therefore turned to detecting discussions of ADEs in social media. Impressive results have been achieved in various attempts to detect ADEs. In a high-stakes domain such as medicine, however, an in-depth evaluation of a model's abilities is crucial. We address the issue of thorough performance evaluation in English-language ADE detection with hand-crafted templates for four capabilities: Temporal order, negation, sentiment, and beneficial effect. We find that models with similar performance on held-out test sets have varying results on these capabilities., Comment: Accepted at BioNLP 2024 and Shared Tasks (ACL Workshop)

Deep learning-based approaches for software vulnerability prediction currently mainly rely on the original text of software code as the feature of nodes in the graph of code and thus could learn a representation that is only specific to the code text, rather than the representation that depicts the 'intrinsic' functionality of a program hidden in the text representation. One curse that causes this problem is an infinite number of possibilities to name a variable. In order to lift the curse, in this work we introduce a new type of edge called name dependence, a type of abstract syntax graph based on the name dependence, and an efficient node representation method named 3-property encoding scheme. These techniques will allow us to remove the concrete variable names from code, and facilitate deep learning models to learn the functionality of software hidden in diverse code expressions. The experimental results show that the deep learning models built on these techniques outperform the ones based on existing approaches not only in the prediction of vulnerabilities but also in the memory need. The factor of memory usage reductions of our techniques can be up to the order of 30,000 in comparison to existing approaches.

Lifting exploits symmetries in probabilistic graphical models by using a representative for indistinguishable objects, allowing to carry out query answering more efficiently while maintaining exact answers. In this paper, we investigate how lifting enables us to perform probabilistic inference for factor graphs containing factors whose potentials are unknown. We introduce the Lifting Factor Graphs with Some Unknown Factors (LIFAGU) algorithm to identify symmetric subgraphs in a factor graph containing unknown factors, thereby enabling the transfer of known potentials to unknown potentials to ensure a well-defined semantics and allow for (lifted) probabilistic inference., Comment: Accepted to the Proceedings of the 17th European Conference on Symbolic and Quantitative Approaches to Reasoning with Uncertainty (ECSQARU-23)

Objectives: To present a publicly available torso segmentation network for large epidemiology datasets on volumetric interpolated breath-hold examination (VIBE) images. Materials & Methods: We extracted preliminary segmentations from TotalSegmentator, spine, and body composition networks for VIBE images, then improved them iteratively and retrained a nnUNet network. Using subsets of NAKO (85 subjects) and UK Biobank (16 subjects), we evaluated with Dice-score on a holdout set (12 subjects) and existing organ segmentation approach (1000 subjects), generating 71 semantic segmentation types for VIBE images. We provide an additional network for the vertebra segments 22 individual vertebra types. Results: We achieved an average Dice score of 0.89 +- 0.07 overall 71 segmentation labels. We scored > 0.90 Dice-score on the abdominal organs except for the pancreas with a Dice of 0.70. Conclusion: Our work offers a detailed and refined publicly available full torso segmentation on VIBE images., Comment: https://github.com/robert-graf/TotalVibeSegmentator

Modeling open hole failure of composites is a complex task, consisting in a highly nonlinear response with interacting failure modes. Numerical modeling of this phenomenon has traditionally been based on the finite element method, but requires to tradeoff between high fidelity and computational cost. To mitigate this shortcoming, recent work has leveraged machine learning to predict the strength of open hole composite specimens. Here, we also propose using data-based models but to tackle open hole composite failure from a classification point of view. More specifically, we show how to train surrogate models to learn the ultimate failure envelope of an open hole composite plate under in-plane loading. To achieve this, we solve the classification problem via support vector machine (SVM) and test different classifiers by changing the SVM kernel function. The flexibility of kernel-based SVM also allows us to integrate the recently developed quantum kernels in our algorithm and compare them with the standard radial basis function (RBF) kernel. Finally, thanks to kernel-target alignment optimization, we tune the free parameters of all kernels to best separate safe and failure-inducing loading states. The results show classification accuracies higher than 90% for RBF, especially after alignment, followed closely by the quantum kernel classifiers.

User-generated data sources have gained significance in uncovering Adverse Drug Reactions (ADRs), with an increasing number of discussions occurring in the digital world. However, the existing clinical corpora predominantly revolve around scientific articles in English. This work presents a multilingual corpus of texts concerning ADRs gathered from diverse sources, including patient fora, social media, and clinical reports in German, French, and Japanese. Our corpus contains annotations covering 12 entity types, four attribute types, and 13 relation types. It contributes to the development of real-world multilingual language models for healthcare. We provide statistics to highlight certain challenges associated with the corpus and conduct preliminary experiments resulting in strong baselines for extracting entities and relations between these entities, both within and across languages., Comment: Accepted at LREC-COLING 2024

Generative AI is expected to have a vast, positive impact on education; however, at present, this potential has not yet been demonstrated at scale at university level. In this study, we present first evidence that generative AI can increase the speed of learning substantially in university students. We tested whether using the AI-powered teaching assistant Syntea affected the speed of learning of hundreds of distance learning students across more than 40 courses at the IU International University of Applied Sciences. Our analysis suggests that using Syntea reduced their study time substantially--by about 27\% on average--in the third month after the release of Syntea. Taken together, the magnitude of the effect and the scalability of the approach implicate generative AI as a key lever to significantly improve and accelerate learning by personalisation.

Preservation of private user data is of paramount importance for high Quality of Experience (QoE) and acceptability, particularly with services treating sensitive data, such as IT-based health services. Whereas anonymization techniques were shown to be prone to data re-identification, synthetic data generation has gradually replaced anonymization since it is relatively less time and resource-consuming and more robust to data leakage. Generative Adversarial Networks (GANs) have been used for generating synthetic datasets, especially GAN frameworks adhering to the differential privacy phenomena. This research compares state-of-the-art GAN-based models for synthetic data generation to generate time-series synthetic medical records of dementia patients which can be distributed without privacy concerns. Predictive modeling, autocorrelation, and distribution analysis are used to assess the Quality of Generating (QoG) of the generated data. The privacy preservation of the respective models is assessed by applying membership inference attacks to determine potential data leakage risks. Our experiments indicate the superiority of the privacy-preserving GAN (PPGAN) model over other models regarding privacy preservation while maintaining an acceptable level of QoG. The presented results can support better data protection for medical use cases in the future.

This paper provides an overview of current approaches for solving inverse problems in imaging using variational methods and machine learning. A special focus lies on point estimators and their robustness against adversarial perturbations. In this context results of numerical experiments for a one-dimensional toy problem are provided, showing the robustness of different approaches and empirically verifying theoretical guarantees. Another focus of this review is the exploration of the subspace of data-consistent solutions through explicit guidance to satisfy specific semantic or textural properties.

Siamese encoders such as sentence transformers are among the least understood deep models. Established attribution methods cannot tackle this model class since it compares two inputs rather than processing a single one. To address this gap, we have recently proposed an attribution method specifically for Siamese encoders (M\"oller et al., 2023). However, it requires models to be adjusted and fine-tuned and therefore cannot be directly applied to off-the-shelf models. In this work, we reassess these restrictions and propose (i) a model with exact attribution ability that retains the original model's predictive performance and (ii) a way to compute approximate attributions for off-the-shelf models. We extensively compare approximate and exact attributions and use them to analyze the models' attendance to different linguistic aspects. We gain insights into which syntactic roles Siamese transformers attend to, confirm that they mostly ignore negation, explore how they judge semantically opposite adjectives, and find that they exhibit lexical bias., Comment: Accepted for EACL 2024, St. Julian's, Malta

Interpretability tools that offer explanations in the form of a dialogue have demonstrated their efficacy in enhancing users' understanding (Slack et al., 2023; Shen et al., 2023), as one-off explanations may fall short in providing sufficient information to the user. Current solutions for dialogue-based explanations, however, often require external tools and modules and are not easily transferable to tasks they were not designed for. With LLMCheckup, we present an easily accessible tool that allows users to chat with any state-of-the-art large language model (LLM) about its behavior. We enable LLMs to generate explanations and perform user intent recognition without fine-tuning, by connecting them with a broad spectrum of Explainable AI (XAI) methods, including white-box explainability tools such as feature attributions, and self-explanations (e.g., for rationale generation). LLM-based (self-)explanations are presented as an interactive dialogue that supports follow-up questions and generates suggestions. LLMCheckupprovides tutorials for operations available in the system, catering to individuals with varying levels of expertise in XAI and supporting multiple input modalities. We introduce a new parsing strategy that substantially enhances the user intent recognition accuracy of the LLM. Finally, we showcase LLMCheckup for the tasks of fact checking and commonsense question answering., Comment: Accepted to NAACL 2024 HCI+NLP workshop; camera-ready version

This paper introduces panoptica, a versatile and performance-optimized package designed for computing instance-wise segmentation quality metrics from 2D and 3D segmentation maps. panoptica addresses the limitations of existing metrics and provides a modular framework that complements the original intersection over union-based panoptic quality with other metrics, such as the distance metric Average Symmetric Surface Distance. The package is open-source, implemented in Python, and accompanied by comprehensive documentation and tutorials. panoptica employs a three-step metrics computation process to cover diverse use cases. The efficacy of panoptica is demonstrated on various real-world biomedical datasets, where an instance-wise evaluation is instrumental for an accurate representation of the underlying clinical task. Overall, we envision panoptica as a valuable tool facilitating in-depth evaluation of segmentation methods., Comment: 15 pages, 6 figures, 3 tables

As of today, state-of-the-art activity recognition from wearable sensors relies on algorithms being trained to classify fixed windows of data. In contrast, video-based Human Activity Recognition, known as Temporal Action Localization (TAL), has followed a segment-based prediction approach, localizing activity segments in a timeline of arbitrary length. This paper is the first to systematically demonstrate the applicability of state-of-the-art TAL models for both offline and near-online Human Activity Recognition (HAR) using raw inertial data as well as pre-extracted latent features as input. Offline prediction results show that TAL models are able to outperform popular inertial models on a multitude of HAR benchmark datasets, with improvements reaching as much as 26% in F1-score. We show that by analyzing timelines as a whole, TAL models can produce more coherent segments and achieve higher NULL-class accuracy across all datasets. We demonstrate that TAL is less suited for the immediate classification of small-sized windows of data, yet offers an interesting perspective on inertial-based HAR -- alleviating the need for fixed-size windows and enabling algorithms to recognize activities of arbitrary length. With design choices and training concepts yet to be explored, we argue that TAL architectures could be of significant value to the inertial-based HAR community. The code and data download to reproduce experiments is publicly available via github.com/mariusbock/tal_for_har., Comment: 19 pages, 7 figures, 4 tables

Despite the success of Siamese encoder models such as sentence transformers (ST), little is known about the aspects of inputs they pay attention to. A barrier is that their predictions cannot be attributed to individual features, as they compare two inputs rather than processing a single one. This paper derives a local attribution method for Siamese encoders by generalizing the principle of integrated gradients to models with multiple inputs. The solution takes the form of feature-pair attributions, and can be reduced to a token-token matrix for STs. Our method involves the introduction of integrated Jacobians and inherits the advantageous formal properties of integrated gradients: it accounts for the model's full computation graph and is guaranteed to converge to the actual prediction. A pilot study shows that in an ST few token-pairs can often explain large fractions of predictions, and it focuses on nouns and verbs. For accurate predictions, it however needs to attend to the majority of tokens and parts of speech., Comment: Accepted to EMNLP'23

Permutation matrices play a key role in matching and assignment problems across the fields, especially in computer vision and robotics. However, memory for explicitly representing permutation matrices grows quadratically with the size of the problem, prohibiting large problem instances. In this work, we propose to tackle the curse of dimensionality of large permutation matrices by approximating them using low-rank matrix factorization, followed by a nonlinearity. To this end, we rely on the Kissing number theory to infer the minimal rank required for representing a permutation matrix of a given size, which is significantly smaller than the problem size. This leads to a drastic reduction in computation and memory costs, e.g., up to $3$ orders of magnitude less memory for a problem of size $n=20000$, represented using $8.4\times10^5$ elements in two small matrices instead of using a single huge matrix with $4\times 10^8$ elements. The proposed representation allows for accurate representations of large permutation matrices, which in turn enables handling large problems that would have been infeasible otherwise. We demonstrate the applicability and merits of the proposed approach through a series of experiments on a range of problems that involve predicting permutation matrices, from linear and quadratic assignment to shape matching problems., Comment: 13 pages, 6 figures

Background: Automated segmentation of spinal MR images plays a vital role both scientifically and clinically. However, accurately delineating posterior spine structures presents challenges. Methods: This retrospective study, approved by the ethical committee, involved translating T1w and T2w MR image series into CT images in a total of n=263 pairs of CT/MR series. Landmark-based registration was performed to align image pairs. We compared 2D paired (Pix2Pix, denoising diffusion implicit models (DDIM) image mode, DDIM noise mode) and unpaired (contrastive unpaired translation, SynDiff) image-to-image translation using "peak signal to noise ratio" (PSNR) as quality measure. A publicly available segmentation network segmented the synthesized CT datasets, and Dice scores were evaluated on in-house test sets and the "MRSpineSeg Challenge" volumes. The 2D findings were extended to 3D Pix2Pix and DDIM. Results: 2D paired methods and SynDiff exhibited similar translation performance and Dice scores on paired data. DDIM image mode achieved the highest image quality. SynDiff, Pix2Pix, and DDIM image mode demonstrated similar Dice scores (0.77). For craniocaudal axis rotations, at least two landmarks per vertebra were required for registration. The 3D translation outperformed the 2D approach, resulting in improved Dice scores (0.80) and anatomically accurate segmentations in a higher resolution than the original MR image. Conclusion: Two landmarks per vertebra registration enabled paired image-to-image translation from MR to CT and outperformed all unpaired approaches. The 3D techniques provided anatomically correct segmentations, avoiding underprediction of small structures like the spinous process., Comment: 35 pages, 7 figures, Code and a model weights available https://doi.org/10.5281/zenodo.8221159 and https://doi.org/10.5281/zenodo.8198697

Zero-cost proxies are nowadays frequently studied and used to search for neural architectures. They show an impressive ability to predict the performance of architectures by making use of their untrained weights. These techniques allow for immense search speed-ups. So far the joint search for well-performing and robust architectures has received much less attention in the field of NAS. Therefore, the main focus of zero-cost proxies is the clean accuracy of architectures, whereas the model robustness should play an evenly important part. In this paper, we analyze the ability of common zero-cost proxies to serve as performance predictors for robustness in the popular NAS-Bench-201 search space. We are interested in the single prediction task for robustness and the joint multi-objective of clean and robust accuracy. We further analyze the feature importance of the proxies and show that predicting the robustness makes the prediction task from existing zero-cost proxies more challenging. As a result, the joint consideration of several proxies becomes necessary to predict a model's robustness while the clean accuracy can be regressed from a single such feature., Comment: Accepted at DAGM GCPR 2023

Backdoors pose a serious threat to machine learning, as they can compromise the integrity of security-critical systems, such as self-driving cars. While different defenses have been proposed to address this threat, they all rely on the assumption that the hardware on which the learning models are executed during inference is trusted. In this paper, we challenge this assumption and introduce a backdoor attack that completely resides within a common hardware accelerator for machine learning. Outside of the accelerator, neither the learning model nor the software is manipulated, so that current defenses fail. To make this attack practical, we overcome two challenges: First, as memory on a hardware accelerator is severely limited, we introduce the concept of a minimal backdoor that deviates as little as possible from the original model and is activated by replacing a few model parameters only. Second, we develop a configurable hardware trojan that can be provisioned with the backdoor and performs a replacement only when the specific target model is processed. We demonstrate the practical feasibility of our attack by implanting our hardware trojan into the Xilinx Vitis AI DPU, a commercial machine-learning accelerator. We configure the trojan with a minimal backdoor for a traffic-sign recognition system. The backdoor replaces only 30 (0.069%) model parameters, yet it reliably manipulates the recognition once the input contains a backdoor trigger. Our attack expands the hardware circuit of the accelerator by 0.24% and induces no run-time overhead, rendering a detection hardly possible. Given the complex and highly distributed manufacturing process of current hardware, our work points to a new threat in machine learning that is inaccessible to current security mechanisms and calls for hardware to be manufactured only in fully trusted environments.

The number of papers submitted to academic conferences is steadily rising in many scientific disciplines. To handle this growth, systems for automatic paper-reviewer assignments are increasingly used during the reviewing process. These systems use statistical topic models to characterize the content of submissions and automate the assignment to reviewers. In this paper, we show that this automation can be manipulated using adversarial learning. We propose an attack that adapts a given paper so that it misleads the assignment and selects its own reviewers. Our attack is based on a novel optimization strategy that alternates between the feature space and problem space to realize unobtrusive changes to the paper. To evaluate the feasibility of our attack, we simulate the paper-reviewer assignment of an actual security conference (IEEE S&P) with 165 reviewers on the program committee. Our results show that we can successfully select and remove reviewers without access to the assignment system. Moreover, we demonstrate that the manipulated papers remain plausible and are often indistinguishable from benign submissions., Comment: Accepted at USENIX Security Symposium 2023

The reconstruction of images from their corresponding noisy Radon transform is a typical example of an ill-posed linear inverse problem as arising in the application of computerized tomography (CT). As the (naive) solution does not depend on the measured data continuously, regularization is needed to re-establish a continuous dependence. In this work, we investigate simple, but yet still provably convergent approaches to learning linear regularization methods from data. More specifically, we analyze two approaches: One generic linear regularization that learns how to manipulate the singular values of the linear operator in an extension of our previous work, and one tailored approach in the Fourier domain that is specific to CT-reconstruction. We prove that such approaches become convergent regularization methods as well as the fact that the reconstructions they provide are typically much smoother than the training data they were trained on. Finally, we compare the spectral as well as the Fourier-based approaches for CT-reconstruction numerically, discuss their advantages and disadvantages and investigate the effect of discretization errors at different resolutions.

Modern quantum annealers can find high-quality solutions to combinatorial optimisation objectives given as quadratic unconstrained binary optimisation (QUBO) problems. Unfortunately, obtaining suitable QUBO forms in computer vision remains challenging and currently requires problem-specific analytical derivations. Moreover, such explicit formulations impose tangible constraints on solution encodings. In stark contrast to prior work, this paper proposes to learn QUBO forms from data through gradient backpropagation instead of deriving them. As a result, the solution encodings can be chosen flexibly and compactly. Furthermore, our methodology is general and virtually independent of the specifics of the target problem type. We demonstrate the advantages of learnt QUBOs on the diverse problem types of graph matching, 2D point cloud alignment and 3D rotation estimation. Our results are competitive with the previous quantum state of the art while requiring much fewer logical and physical qubits, enabling our method to scale to larger problems. The code and the new dataset will be open-sourced., Comment: incl. appendix

The powerful modeling capabilities of all-attention-based transformer architectures often cause overfitting and - for natural language processing tasks - lead to an implicitly learned internal language model in the autoregressive transformer decoder complicating the integration of external language models. In this paper, we explore relaxed attention, a simple and easy-to-implement smoothing of the attention weights, yielding a two-fold improvement to the general transformer architecture: First, relaxed attention provides regularization when applied to the self-attention layers in the encoder. Second, we show that it naturally supports the integration of an external language model as it suppresses the implicitly learned internal language model by relaxing the cross attention in the decoder. We demonstrate the benefit of relaxed attention across several tasks with clear improvement in combination with recent benchmark approaches. Specifically, we exceed the former state-of-the-art performance of 26.90% word error rate on the largest public lip-reading LRS3 benchmark with a word error rate of 26.31%, as well as we achieve a top-performing BLEU score of 37.67 on the IWSLT14 (DE$\rightarrow$EN) machine translation task without external language models and virtually no additional model parameters. Code and models will be made publicly available.

In this work, we present the first corpus for German Adverse Drug Reaction (ADR) detection in patient-generated content. The data consists of 4,169 binary annotated documents from a German patient forum, where users talk about health issues and get advice from medical doctors. As is common in social media data in this domain, the class labels of the corpus are very imbalanced. This and a high topic imbalance make it a very challenging dataset, since often, the same symptom can have several causes and is not always related to a medication intake. We aim to encourage further multi-lingual efforts in the domain of ADR detection and provide preliminary experiments for binary classification using different methods of zero- and few-shot learning based on a multi-lingual model. When fine-tuning XLM-RoBERTa first on English patient forum data and then on the new German data, we achieve an F1-score of 37.52 for the positive class. We make the dataset and models publicly available for the community., Comment: Accepted at LREC 2022

Text readability assessment has a wide range of applications for different target people, from language learners to people with disabilities. The fast pace of textual content production on the web makes it impossible to measure text complexity without the benefit of machine learning and natural language processing techniques. Although various research addressed the readability assessment of English text in recent years, there is still room for improvement of the models for other languages. In this paper, we proposed a new model for text complexity assessment for German text based on transfer learning. Our results show that the model outperforms more classical solutions based on linguistic features extraction from input text. The best model is based on the BERT pre-trained language model achieved the Root Mean Square Error (RMSE) of 0.483.

The human-centric explainable artificial intelligence (HCXAI) community has raised the need for framing the explanation process as a conversation between human and machine. In this position paper, we establish desiderata for Mediators, text-based conversational agents which are capable of explaining the behavior of neural models interactively using natural language. From the perspective of natural language processing (NLP) research, we engineer a blueprint of such a Mediator for the task of sentiment analysis and assess how far along current research is on the path towards dialogue-based explanations., Comment: Accepted to IJCAI-ECAI 2022 Workshop on Explainable Artificial Intelligence (XAI)

Scientific publications about machine learning in healthcare are often about implementing novel methods and boosting the performance - at least from a computer science perspective. However, beyond such often short-lived improvements, much more needs to be taken into consideration if we want to arrive at a sustainable progress in healthcare. What does it take to actually implement such a system, make it usable for the domain expert, and possibly bring it into practical usage? Targeted at Computer Scientists, this work presents a multidisciplinary view on machine learning in medical decision support systems and covers information technology, medical, as well as ethical aspects. Along with an implemented risk prediction system in nephrology, challenges and lessons learned in a pilot project are presented., Comment: Paper currently under review

Recent studies show that deep learning (DL) based MRI reconstruction outperforms conventional methods, such as parallel imaging and compressed sensing (CS), in multiple applications. Unlike CS that is typically implemented with pre-determined linear representations for regularization, DL inherently uses a non-linear representation learned from a large database. Another line of work uses transform learning (TL) to bridge the gap between these two approaches by learning linear representations from data. In this work, we combine ideas from CS, TL and DL reconstructions to learn deep linear convolutional transforms as part of an algorithm unrolling approach. Using end-to-end training, our results show that the proposed technique can reconstruct MR images to a level comparable to DL methods, while supporting uniform undersampling patterns unlike conventional CS methods. Our proposed method relies on convex sparse image reconstruction with linear representation at inference time, which may be beneficial for characterizing robustness, stability and generalizability.

Despite their effective use in various fields, many aspects of neural networks are poorly understood. One important way to investigate the characteristics of neural networks is to explore the loss landscape. However, most models produce a high-dimensional non-convex landscape which is difficult to visualize. We discuss and extend existing visualization methods based on 1D- and 2D slicing with a novel method that approximates the actual loss landscape geometry by using charts with interpretable axes. Based on the assumption that observations on small neural networks can generalize to more complex systems and provide us with helpful insights, we focus on small models in the range of a few dozen weights, which enables computationally cheap experiments and the use of an interactive dashboard. We observe symmetries around the zero vector, the influence of different layers on the global landscape, the different weight sensitivities around a minimizer, and how gradient descent navigates high-loss obstacles. The user study resulted in an average SUS (System Usability Scale) score with suggestions for improvement and opened up a number of possible application scenarios, such as autoencoders and ensemble networks., Comment: 10 pages, 18 figures, under review

Neural fields have gained significant attention in the computer vision community due to their excellent performance in novel view synthesis, geometry reconstruction, and generative modeling. Some of their advantages are a sound theoretic foundation and an easy implementation in current deep learning frameworks. While neural fields have been applied to signals on manifolds, e.g., for texture reconstruction, their representation has been limited to extrinsically embedding the shape into Euclidean space. The extrinsic embedding ignores known intrinsic manifold properties and is inflexible wrt. transfer of the learned function. To overcome these limitations, this work introduces intrinsic neural fields, a novel and versatile representation for neural fields on manifolds. Intrinsic neural fields combine the advantages of neural fields with the spectral properties of the Laplace-Beltrami operator. We show theoretically that intrinsic neural fields inherit many desirable properties of the extrinsic neural field framework but exhibit additional intrinsic qualities, like isometry invariance. In experiments, we show intrinsic neural fields can reconstruct high-fidelity textures from images with state-of-the-art quality and are robust to the discretization of the underlying manifold. We demonstrate the versatility of intrinsic neural fields by tackling various applications: texture transfer between deformed shapes & different shapes, texture reconstruction from real-world images with view dependence, and discretization-agnostic learning on meshes and point clouds.

Vocabulary learning is vital to foreign language learning. Correct and adequate feedback is essential to successful and satisfying vocabulary training. However, many vocabulary and language evaluation systems perform on simple rules and do not account for real-life user learning data. This work introduces Multi-Language Vocabulary Evaluation Data Set (MuLVE), a data set consisting of vocabulary cards and real-life user answers, labeled indicating whether the user answer is correct or incorrect. The data source is user learning data from the Phase6 vocabulary trainer. The data set contains vocabulary questions in German and English, Spanish, and French as target language and is available in four different variations regarding pre-processing and deduplication. We experiment to fine-tune pre-trained BERT language models on the downstream task of vocabulary evaluation with the proposed MuLVE data set. The results provide outstanding results of > 95.5 accuracy and F2-score. The data set is available on the European Language Grid., Comment: Submitted to LREC 2022

The detection of hate speech online has become an important task, as offensive language such as hurtful, obscene and insulting content can harm marginalized people or groups. This paper presents TU Berlin team experiments and results on the task 1A and 1B of the shared task on hate speech and offensive content identification in Indo-European languages 2021. The success of different Natural Language Processing models is evaluated for the respective subtasks throughout the competition. We tested different models based on recurrent neural networks in word and character levels and transfer learning approaches based on Bert on the provided dataset by the competition. Among the tested models that have been used for the experiments, the transfer learning-based models achieved the best results in both subtasks., Comment: Accepted at FIRE 2021 - Hate Speech and offensive content detection (HASOC) Track

Estimating quality of transmitted speech is known to be a non-trivial task. While traditionally, test participants are asked to rate the quality of samples; nowadays, automated methods are available. These methods can be divided into: 1) intrusive models, which use both, the original and the degraded signals, and 2) non-intrusive models, which only require the degraded signal. Recently, non-intrusive models based on neural networks showed to outperform signal processing based models. However, the advantages of deep learning based models come with the cost of being more challenging to interpret. To get more insight into the prediction models the non-intrusive speech quality prediction model NISQA is analyzed in this paper. NISQA is composed of a convolutional neural network (CNN) and a recurrent neural network (RNN). The task of the CNN is to compute relevant features for the speech quality prediction on a frame level, while the RNN models time-dependencies between the individual speech frames. Different explanation algorithms are used to understand the automatically learned features of the CNN. In this way, several interpretable features could be identified, such as the sensitivity to noise or strong interruptions. On the other hand, it was found that multiple features carry redundant information., Comment: 4 pages, 6 figures, In Proceedings of the DAGA 2021 (the annual conference of the German Acoustical Society, DEGA)

Wikidata is a frequently updated, community-driven, and multilingual knowledge graph. Hence, Wikidata is an attractive basis for Entity Linking, which is evident by the recent increase in published papers. This survey focuses on four subjects: (1) Which Wikidata Entity Linking datasets exist, how widely used are they and how are they constructed? (2) Do the characteristics of Wikidata matter for the design of Entity Linking datasets and if so, how? (3) How do current Entity Linking approaches exploit the specific characteristics of Wikidata? (4) Which Wikidata characteristics are unexploited by existing Entity Linking approaches? This survey reveals that current Wikidata-specific Entity Linking datasets do not differ in their annotation scheme from schemes for other knowledge graphs like DBpedia. Thus, the potential for multilingual and time-dependent datasets, naturally suited for Wikidata, is not lifted. Furthermore, we show that most Entity Linking approaches use Wikidata in the same way as any other knowledge graph missing the chance to leverage Wikidata-specific characteristics to increase quality. Almost all approaches employ specific properties like labels and sometimes descriptions but ignore characteristics such as the hyper-relational structure. Hence, there is still room for improvement, for example, by including hyper-relational graph embeddings or type information. Many approaches also include information from Wikipedia, which is easily combinable with Wikidata and provides valuable textual information, which Wikidata lacks., Comment: Disclaimer: Cedric M\"oller, Jens Lehmann, Ricardo Usbeck, 2021. The definitive, peer reviewed and edited version of this article is published in the Semantic Web Journal, Special issue: Latest Advancements in Linguistic 3 Linked Data, 2021

We describe how the Fink broker early supernova Ia classifier optimizes its ML classifications by employing an active learning (AL) strategy. We demonstrate the feasibility of implementation of such strategies in the current Zwicky Transient Facility (ZTF) public alert data stream. We compare the performance of two AL strategies: uncertainty sampling and random sampling. Our pipeline consists of 3 stages: feature extraction, classification and learning strategy. Starting from an initial sample of 10 alerts (5 SN Ia and 5 non-Ia), we let the algorithm identify which alert should be added to the training sample. The system is allowed to evolve through 300 iterations. Our data set consists of 23 840 alerts from the ZTF with confirmed classification via cross-match with SIMBAD database and the Transient name server (TNS), 1 600 of which were SNe Ia (1 021 unique objects). The data configuration, after the learning cycle was completed, consists of 310 alerts for training and 23 530 for testing. Averaging over 100 realizations, the classifier achieved 89% purity and 54% efficiency. From 01/November/2020 to 31/October/2021 Fink has applied its early supernova Ia module to the ZTF stream and communicated promising SN Ia candidates to the TNS. From the 535 spectroscopically classified Fink candidates, 459 (86%) were proven to be SNe Ia. Our results confirm the effectiveness of active learning strategies for guiding the construction of optimal training samples for astronomical classifiers. It demonstrates in real data that the performance of learning algorithms can be highly improved without the need of extra computational resources or overwhelmingly large training samples. This is, to our knowledge, the first application of AL to real alerts data., Comment: accepted for publication in Astronomy and Astrophysics