Search

Autonomous Driving Systems (ADSs) rely on Deep Neural Networks, allowing vehicles to navigate complex, open environments. However, the unpredictability of these scenarios highlights the need for rigorous system-level testing to ensure safety, a task usually performed with a simulator in the loop. Though one important goal of such testing is to detect safety violations, there are many undesirable system behaviors, that may not immediately lead to violations, that testing should also be focusing on, thus detecting more subtle problems and enabling a finer-grained analysis. This paper introduces Cooperative Co-evolutionary MEtamorphic test Generator for Autonomous systems (CoCoMEGA), a novel automated testing framework aimed at advancing system-level safety assessments of ADSs. CoCoMEGA combines Metamorphic Testing (MT) with a search-based approach utilizing Cooperative Co-Evolutionary Algorithms (CCEA) to efficiently generate a diverse set of test cases. CoCoMEGA emphasizes the identification of test scenarios that present undesirable system behavior, that may eventually lead to safety violations, captured by Metamorphic Relations (MRs). When evaluated within the CARLA simulation environment on the Interfuser ADS, CoCoMEGA consistently outperforms baseline methods, demonstrating enhanced effectiveness and efficiency in generating severe, diverse MR violations and achieving broader exploration of the test space. These results underscore CoCoMEGA as a promising, more scalable solution to the inherent challenges in ADS testing with a simulator in the loop. Future research directions may include extending the approach to additional simulation platforms, applying it to other complex systems, and exploring methods for further improving testing efficiency such as surrogate modeling., Comment: 24 pages, 6 figures, to be published in IEEE Transactions on Software Engineering

Cyber-Physical Systems (CPSs) software is used to enforce desired behaviours on physical systems. To test the interaction between the CPS software and the system's physics, engineers provide traces of desired physical states and observe traces of the actual physical states. CPS requirements describe how closely the actual physical traces should track the desired traces. These requirements are typically defined for specific, simple input traces such as step or ramp sequences, and thus are not applicable to arbitrary inputs. This limits the availability of oracles for CPSs. Our recent work proposes an approach to testing CPS using control-theoretical design assumptions instead of requirements. This approach circumvents the oracle problem by leveraging the control-theoretical guarantees that are provided when the design assumptions are satisfied. To address the test case generation and oracle problems, researchers have proposed metamorphic testing, which is based on the study of relations across tests, i.e., metamorphic relations (MRs). In this work, we define MRs based on the design assumptions and explore combinations of these MRs using genetic programming to generate CPS test cases. This enables the generation of CPS input traces with potentially arbitrary shapes, together with associated expected output traces. We use the deviation from the expected output traces to guide the generation of input traces that falsify the MRs. Our experiment results show that the MR-falsification provides engineers with new information, helping them identify passed and failed test cases. Furthermore, we show that the generation of traces that falsify the MRs is a non-trivial problem, which is successfully addressed by our genetic search.

Controllers for software-defined networks (SDNs) are centralised software components that enable advanced network functionalities, such as dynamic traffic engineering and network virtualisation. However, these functionalities increase the complexity of SDN controllers, making thorough testing crucial. SDN controllers are stateful, interacting with multiple network devices through sequences of control messages. Identifying stateful failures in an SDN controller is challenging due to the infinite possible sequences of control messages, which result in an unbounded number of stateful interactions between the controller and network devices. In this article, we propose SeqFuzzSDN, a learning-guided fuzzing method for testing stateful SDN controllers. SeqFuzzSDN aims to (1) efficiently explore the state space of the SDN controller under test, (2) generate effective and diverse tests (i.e., control message sequences) to uncover failures, and (3) infer accurate failure-inducing models that characterise the message sequences leading to failures. In addition, we compare SeqFuzzSDN with three extensions of state-of-the-art (SOTA) methods for fuzzing SDNs. Our findings show that, compared to the extended SOTA methods, SeqFuzzSDN (1) generates more diverse message sequences that lead to failures within the same time budget, and (2) produces more accurate failure-inducing models, significantly outperforming the other extended SOTA methods in terms of sensitivity.

Deep Neural Networks (DNNs) are increasingly deployed across applications. However, ensuring their reliability remains a challenge, and in many situations, alternative models with similar functionality and accuracy are available. Traditional accuracy-based evaluations often fail to capture behavioral differences between models, especially with limited test datasets, making it difficult to select or combine models effectively. Differential testing addresses this by generating test inputs that expose discrepancies in DNN model behavior. However, existing approaches face significant limitations: many rely on model internals or are constrained by available seed inputs. To address these challenges, we propose DiffGAN, a black-box test image generation approach for differential testing of DNN models. DiffGAN leverages a Generative Adversarial Network (GAN) and the Non-dominated Sorting Genetic Algorithm II to generate diverse and valid triggering inputs that reveal behavioral discrepancies between models. DiffGAN employs two custom fitness functions, focusing on diversity and divergence, to guide the exploration of the GAN input space and identify discrepancies between models' outputs. By strategically searching this space, DiffGAN generates inputs with specific features that trigger differences in model behavior. DiffGAN is black-box, making it applicable in more situations. We evaluate DiffGAN on eight DNN model pairs trained on widely used image datasets. Our results show DiffGAN significantly outperforms a SOTA baseline, generating four times more triggering inputs, with greater diversity and validity, within the same budget. Additionally, the generated inputs improve the accuracy of a machine learning-based model selection mechanism, which selects the best-performing model based on input characteristics and can serve as a smart output voting mechanism when using alternative models.

The structural dynamics of biological macromolecules, such as proteins, DNA/RNA, or their complexes, are strongly influenced by protonation changes of their typically many titratable groups, which explains their pH sensitivity. In turn, conformational and environmental changes in the biomolecule affect the protonation state of these groups. With a few exceptions, conventional force field-based molecular dynamics (MD) simulations do not account for these effects, nor do they allow for coupling to a pH buffer. The $\lambda$-dynamics method implements this coupling and thus allows for MD simulations at constant pH. It uses separate Hamiltonians for the protonated and deprotonated states of each titratable group, with a $\lambda$ variable that continuously interpolates between them. However, rigorous implementations of Hamiltonian Interpolation (HI) $\lambda$-dynamics are prohibitively slow when used with Particle Mesh Ewald (PME). To circumvent this problem, it has been proposed to interpolate the charges instead of the Hamiltonians (QI). Here, we propose a rigorous yet efficient Multipole-Accelerated Hamiltonian Interpolation (MAHI) method to perform $\lambda$-dynamics in GROMACS. Starting from a charge-scaled Hamiltonian, precomputed with the Fast Multipole Method (FMM) or with PME, the correct HI forces are calculated with negligible computational overhead. We compare HI with QI and show that HI leads to more frequent transitions between protonation states, resulting in better sampling and accuracy. Our performance benchmarks show that introducing, e.g., 512 titratable sites to a one million atom MD system increases runtime by less than 20% compared to a regular FMM-based simulation. We have integrated the scheme into our GPU-FMM code for the simulation software GROMACS, allowing an easy and effortless transition from standard force field simulations to constant pH simulations.

The structural dynamics of biological macromolecules, such as proteins, DNA/RNA, or complexes thereof, are strongly influenced by protonation changes of their typically many titratable groups, which explains their sensitivity to pH changes. Conversely, conformational and environmental changes of the biomolecule affect the protonation state of these groups. With few exceptions, conventional force field-based molecular dynamics (MD) simulations do not account for these effects, nor do they allow for coupling to a pH buffer. Here we present a GROMACS implementation of a rigorous Hamiltonian interpolation $\lambda$-dynamics constant pH method, which rests on GPU-accelerated Fast Multipole Method (FMM) electrostatics. Our implementation supports both CHARMM36m and Amber99sb*-ILDN force fields and is largely automated to enable seamless switching from regular MD to constant pH MD, involving minimal changes to the input files. Here, the first of two companion papers describes the underlying constant pH protocol and sample applications to several prototypical benchmark systems such as cardiotoxin V, lysozyme, and staphylococcal nuclease. Enhanced convergence is achieved through a new dynamic barrier height optimization method, and high p$K_a$ accuracy is demonstrated. We use Functional Mode Analysis and Mutual Information to explore the complex intra- and intermolecular couplings between the protonation states of titratable groups as well as those between protonation states and conformational dynamics. We identify striking conformation-dependent p$K_a$ variations and unexpected inter-residue couplings. Conformation-protonation coupling is identified as a primary cause of the slow protonation convergence notorious to constant pH simulations involving multiple titratable groups, suggesting enhanced sampling methods to accelerate convergence.

We present the first combinatorial proof of the Graham-Pollak Formula for the determinant of the distance matrix of a tree, via sign-reversing involutions and the Lindstr\"om-Gessel-Viennot Lemma. Our approach provides a cohesive and unified framework for the understanding of the existing generalizations and $q$-analogues of the Graham-Pollak Formula, and facilitates the derivation of a natural simultaneous generalizations for them.

In safety-critical systems (e.g., autonomous vehicles and robots), Deep Neural Networks (DNNs) are becoming a key component for computer vision tasks, particularly semantic segmentation. Further, since the DNN behavior cannot be assessed through code inspection and analysis, test automation has become an essential activity to gain confidence in the reliability of DNNs. Unfortunately, state-of-the-art automated testing solutions largely rely on simulators, whose fidelity is always imperfect, thus affecting the validity of test results. To address such limitations, we propose to combine meta-heuristic search, used to explore the input space using simulators, with Generative Adversarial Networks (GANs), to transform the data generated by simulators into realistic input images. Such images can be used both to assess the DNN performance and to retrain the DNN more effectively. We applied our approach to a state-of-the-art DNN performing semantic segmentation and demonstrated that it outperforms a state-of-the-art GAN-based testing solution and several baselines. Specifically, it leads to the largest number of diverse images leading to the worst DNN performance. Further, the images generated with our approach, lead to the highest improvement in DNN performance when used for retraining. In conclusion, we suggest to always integrate GAN components when performing search-driven, simulator-based testing., Comment: 14 pages, 3 figures, 7 tables

Log-based anomaly detection has been widely studied in the literature as a way to increase the dependability of software-intensive systems. In reality, logs can be unstable due to changes made to the software during its evolution. This, in turn, degrades the performance of downstream log analysis activities, such as anomaly detection. The critical challenge in detecting anomalies on these unstable logs is the lack of information about the new logs, due to insufficient log data from new software versions. The application of Large Language Models (LLMs) to many software engineering tasks has revolutionized various domains. In this paper, we report on an experimental comparison of a fine-tuned LLM and alternative models for anomaly detection on unstable logs. The main motivation is that the pre-training of LLMs on vast datasets may enable a robust understanding of diverse patterns and contextual information, which can be leveraged to mitigate the data insufficiency issue in the context of software evolution. Our experimental results on the two-version dataset of LOGEVOL-Hadoop show that the fine-tuned LLM (GPT-3) fares slightly better than supervised baselines when evaluated on unstable logs. The difference between GPT-3 and other supervised approaches tends to become more significant as the degree of changes in log sequences increases. However, it is unclear whether the difference is practically significant in all cases. Lastly, our comparison of prompt engineering (with GPT-4) and fine-tuning reveals that the latter provides significantly superior performance on both stable and unstable logs, offering valuable insights into the effective utilization of LLMs in this domain.

Mutation testing consists of generating test cases that detect faults injected into software (generating mutants) which its original test suite could not. By running such an augmented set of test cases, it may discover actual faults that may have gone unnoticed with the original test suite. It is thus a desired practice for embedded software running in safety-critical cyber-physical systems (CPS). Unfortunately, the state-of-the-art tool targeting C, a typical language for CPS software, relies on symbolic execution, whose limitations often prevent its application. MOTIF overcomes such limitations by leveraging grey-box fuzzing tools to generate unit test cases in C that detect injected faults in mutants. Indeed, fuzzing tools automatically generate inputs by exercising the compiled version of the software under test guided by coverage feedback, thus overcoming the limitations of symbolic execution. Our empirical assessment has shown that it detects more faults than symbolic execution (i.e., up to 47 percentage points), when the latter is applicable., Comment: ICST2024. arXiv admin note: text overlap with arXiv:2308.07949

In learning-enabled autonomous systems, safety monitoring of learned components is crucial to ensure their outputs do not lead to system safety violations, given the operational context of the system. However, developing a safety monitor for practical deployment in real-world applications is challenging. This is due to limited access to internal workings and training data of the learned component. Furthermore, safety monitors should predict safety violations with low latency, while consuming a reasonable amount of computation. To address the challenges, we propose a safety monitoring method based on probabilistic time series forecasting. Given the learned component outputs and an operational context, we empirically investigate different Deep Learning (DL)-based probabilistic forecasting to predict the objective measure capturing the satisfaction or violation of a safety requirement (safety metric). We empirically evaluate safety metric and violation prediction accuracy, and inference latency and resource usage of four state-of-the-art models, with varying horizons, using autonomous aviation and autonomous driving case studies. Our results suggest that probabilistic forecasting of safety metrics, given learned component outputs and scenarios, is effective for safety monitoring. Furthermore, for both case studies, Temporal Fusion Transformer (TFT) was the most accurate model for predicting imminent safety violations, with acceptable latency and resource consumption.

Solar radio spikes are short lived, narrow bandwidth features in low frequency solar radio observations. The timing of their occurrence and the number of spikes in a given observation is often unpredictable. The high temporal and frequency of resolution of modern radio telescopes such as NenuFAR mean that manually identifying radio spikes is an arduous task. Machine learning approaches to data exploration in solar radio data is on the rise. Here we describe a convolutional neural network to identify the per pixel location of radio spikes as well as determine some simple characteristics of duration, spectral width and drift rate. The model, which we call SpikeNet, was trained using an Nvidia Tesla T4 16GB GPU with ~100000 sample spikes in a total time of 2.2 hours. The segmentation performs well with an intersection over union in the test set of ~0.85. The root mean squared error for predicted spike properties is of the order of 23%. Applying the algorithm to unlabelled data successfully generates segmentation masks although the accuracy of the predicted properties is less reliable, particularly when more than one spike is present in the same 64 X 64 pixel time-frequency range. We have successfully demonstrated that our convolutional neural network can locate and characterise solar radio spikes in a number of seconds compared to the weeks it would take for manual identification., Comment: 7 pages, 7 figures

Although the security testing of Web systems can be automated by generating crafted inputs, solutions to automate the test oracle, i.e., vulnerability detection, remain difficult to apply in practice. Specifically, though previous work has demonstrated the potential of metamorphic testing, security failures can be determined by metamorphic relations that turn valid inputs into malicious inputs, metamorphic relations are typically executed on a large set of inputs, which is time-consuming and thus makes metamorphic testing impractical. We propose AIM, an approach that automatically selects inputs to reduce testing costs while preserving vulnerability detection capabilities. AIM includes a clustering-based black-box approach, to identify similar inputs based on their security properties. It also relies on a novel genetic algorithm to efficiently select diverse inputs while minimizing their total cost. Further, it contains a problem-reduction component to reduce the search space and speed up the minimization process. We evaluated the effectiveness of AIM on two well-known Web systems, Jenkins and Joomla, with documented vulnerabilities. We compared AIM's results with four baselines involving standard search approaches. Overall, AIM reduced metamorphic testing time by 84% for Jenkins and 82% for Joomla, while preserving the same level of vulnerability detection. Furthermore, AIM significantly outperformed all the considered baselines regarding vulnerability coverage.

We explore the AI2050 "hard problems" that block the promise of AI and cause AI risks: (1) developing general capabilities of the systems; (2) assuring the performance of AI systems and their training processes; (3) aligning system goals with human goals; (4) enabling great applications of AI in real life; (5) addressing economic disruptions; (6) ensuring the participation of all; (7) at the same time ensuring socially responsible deployment; (8) addressing any geopolitical disruptions that AI causes; (9) promoting sound governance of the technology; and (10) managing the philosophical disruptions for humans living in the age of AI. For each problem, we outline the area, identify significant recent work, and suggest ways forward. [Note: this paper reviews literature through January 2023.], Comment: 75 + 19 pages

Ensuring the quality of software systems through testing is essential, yet maintaining test cases poses significant challenges and costs. The need for frequent updates to align with the evolving system under test often entails high complexity and cost for maintaining these test cases. Further, unrepaired broken test cases can degrade test suite quality and disrupt the software development process, wasting developers' time. To address this challenge, we present TaRGet (Test Repair GEneraTor), a novel approach leveraging pre-trained code language models for automated test case repair. TaRGet treats test repair as a language translation task, employing a two-step process to fine-tune a language model based on essential context data characterizing the test breakage. To evaluate our approach, we introduce TaRBench, a comprehensive benchmark we developed covering 45,373 broken test repairs across 59 open-source projects. Our results demonstrate TaRGet's effectiveness, achieving a 66.1% exact match accuracy. Furthermore, our study examines the effectiveness of TaRGet across different test repair scenarios. We provide a practical guide to predict situations where the generated test repairs might be less reliable. We also explore whether project-specific data is always necessary for fine-tuning and if our approach can be effective on new projects., Comment: 40 pages, 18 figures

Solar radio bursts are some of the brightest emissions at radio frequencies in the solar system. The emission mechanisms that generate these bursts offer a remote insight into physical processes in solar coronal plasma, while fine spectral features hint at its underlying turbulent nature. During radio noise storms many hundreds of solar radio bursts can occur over the course of a few hours. Identifying and classifying solar radio bursts is often done manually although a number of automatic algorithms have been produced for this purpose. The use of machine learning algorithms for image segmentation and classification is well established and has shown promising results in the case of identifying Type II and Type III solar radio bursts. Here we present the results of a convolutional neural network applied to dynamic spectra of NenuFAR solar observations. We highlight some initial success in segmenting radio bursts from the background spectra and outline the steps necessary for burst classification., Comment: 11 pages, 4 figures, accepted for publication in PRE9 conference proceedings

This paper presents our recent initiatives to foster the discoverability of new releases on the music streaming service Deezer. After introducing our search and recommendation features dedicated to new releases, we outline our shift from editorial to personalized release suggestions using cold start embeddings and contextual bandits. Backed by online experiments, we discuss the advantages of this shift in terms of recommendation quality and exposure of new releases on the service., Comment: Accepted for presentation as an "Industry Talk" at the 46th European Conference on Information Retrieval (ECIR 2024)

Modern software has been an integral part of everyday activities in many disciplines and application contexts. Introducing intelligent automation by leveraging artificial intelligence (AI) led to break-throughs in many fields. The effectiveness of AI can be attributed to several factors, among which is the increasing availability of data. Regulations such as the general data protection regulation (GDPR) in the European Union (EU) are introduced to ensure the protection of personal data. Software systems that collect, process, or share personal data are subject to compliance with such regulations. Developing compliant software depends heavily on addressing legal requirements stipulated in applicable regulations, a central activity in the requirements engineering (RE) phase of the software development process. RE is concerned with specifying and maintaining requirements of a system-to-be, including legal requirements. Legal agreements which describe the policies organizations implement for processing personal data can provide an additional source to regulations for eliciting legal requirements. In this chapter, we explore a variety of methods for analyzing legal requirements and exemplify them on GDPR. Specifically, we describe possible alternatives for creating machine-analyzable representations from regulations, survey the existing automated means for enabling compliance verification against regulations, and further reflect on the current challenges of legal requirements analysis.

Data entry forms use completeness requirements to specify the fields that are required or optional to fill for collecting necessary information from different types of users. However, some required fields may not be applicable for certain types of users anymore. Nevertheless, they may still be incorrectly marked as required in the form; we call such fields obsolete required fields. Since obsolete required fields usually have not-null validation checks before submitting the form, users have to enter meaningless values in such fields in order to complete the form submission. These meaningless values threaten the quality of the filled data. To avoid users filling meaningless values, existing techniques usually rely on manually written rules to identify the obsolete required fields and relax their completeness requirements. However, these techniques are ineffective and costly. In this paper, we propose LACQUER, a learning-based automated approach for relaxing the completeness requirements of data entry forms. LACQUER builds Bayesian Network models to automatically learn conditions under which users had to fill meaningless values. To improve its learning ability, LACQUER identifies the cases where a required field is only applicable for a small group of users, and uses SMOTE, an oversampling technique, to generate more instances on such fields for effectively mining dependencies on them. Our experimental results show that LACQUER can accurately relax the completeness requirements of required fields in data entry forms with precision values ranging between 0.76 and 0.90 on different datasets. LACQUER can prevent users from filling 20% to 64% of meaningless values, with negative predictive values between 0.72 and 0.91. Furthermore, LACQUER is efficient; it takes at most 839 ms to predict the completeness requirement of an instance., Comment: Accepted for publication by ACM Transactions on Software Engineering and Methodology

Proton magnetic resonance spectroscopic imaging (1H-MRSI) is a powerful tool that enables the multidimensional non-invasive mapping of the neurochemical profile at high-resolution over the entire brain. The constant demand for higher spatial resolution in 1H-MRSI led to increased interest in post-processing-based denoising methods aimed at reducing noise variance. The aim of the present study was to implement two noise-reduction techniques, the Marchenko-Pastur principal component analysis (MP-PCA) based denoising and the low-rank total generalized variation (LR-TGV) reconstruction, and to test their potential and impact on preclinical 14.1T fast in vivo 1H-FID-MRSI datasets. Since there is no known ground truth for in vivo metabolite maps, additional evaluations of the performance of both noise-reduction strategies were conducted using Monte-Carlo simulations. Results showed that both denoising techniques increased the apparent signal-to-noise ratio SNR while preserving noise properties in each spectrum for both in vivo and Monte-Carlo datasets. Relative metabolite concentrations were not significantly altered by either methods and brain regional differences were preserved in both synthetic and in vivo datasets. Increased precision of metabolite estimates was observed for the two methods, with inconsistencies noted on lower concentrated metabolites. Our study provided a framework on how to evaluate the performance of MP-PCA and LR-TGV methods for preclinical 1H-FID MRSI data at 14.1T. While gains in apparent SNR and precision were observed, concentration estimations ought to be treated with care especially for low-concentrated metabolites., Comment: Brayan Alves and Dunja Simicic are joint first authors. Currently in revision for NMR in Biomedicine

Mutation testing can help reduce the risks of releasing faulty software. For such reason, it is a desired practice for the development of embedded software running in safety-critical cyber-physical systems (CPS). Unfortunately, state-of-the-art test data generation techniques for mutation testing of C and C++ software, two typical languages for CPS software, rely on symbolic execution, whose limitations often prevent its application (e.g., it cannot test black-box components). We propose a mutation testing approach that leverages fuzz testing, which has proved effective with C and C++ software. Fuzz testing automatically generates diverse test inputs that exercise program branches in a varied number of ways and, therefore, exercise statements in different program states, thus maximizing the likelihood of killing mutants, our objective. We performed an empirical assessment of our approach with software components used in satellite systems currently in orbit. Our empirical evaluation shows that mutation testing based on fuzz testing kills a significantly higher proportion of live mutants than symbolic execution (i.e., up to an additional 47 percentage points). Further, when symbolic execution cannot be applied, fuzz testing provides significant benefits (i.e., up to 41% mutants killed). Our study is the first one comparing fuzz testing and symbolic execution for mutation testing; our results provide guidance towards the development of fuzz testing tools dedicated to mutation testing., Comment: This article is the camera-ready version for ASE 2023

Although App updates are frequent and software engineers would like to verify updated features only, automated testing techniques verify entire Apps and are thus wasting resources. We present Continuous Adaptation of Learned Models (CALM), an automated App testing approach that efficiently test App updates by adapting App models learned when automatically testing previous App versions. CALM focuses on functional testing. Since functional correctness can be mainly verified through the visual inspection of App screens, CALM minimizes the number of App screens to be visualized by software testers while maximizing the percentage of updated methods and instructions exercised. Our empirical evaluation shows that CALM exercises a significantly higher proportion of updated methods and instructions than six state-of-the-art approaches, for the same maximum number of App screens to be visually inspected. Further, in common update scenarios, where only a small fraction of methods are updated, CALM is even quicker to outperform all competing approaches in a more significant way.

Deep Reinforcement Learning (DRL) has made significant advancements in various fields, such as autonomous driving, healthcare, and robotics, by enabling agents to learn optimal policies through interactions with their environments. However, the application of DRL in safety-critical domains presents challenges, particularly concerning the safety of the learned policies. DRL agents, which are focused on maximizing rewards, may select unsafe actions, leading to safety violations. Runtime safety monitoring is thus essential to ensure the safe operation of these agents, especially in unpredictable and dynamic environments. This paper introduces SMARLA, a black-box safety monitoring approach specifically designed for DRL agents. SMARLA utilizes machine learning to predict safety violations by observing the agent's behavior during execution. The approach is based on Q-values, which reflect the expected reward for taking actions in specific states. SMARLA employs state abstraction to reduce the complexity of the state space, enhancing the predictive capabilities of the monitoring model. Such abstraction enables the early detection of unsafe states, allowing for the implementation of corrective and preventive measures before incidents occur. We quantitatively and qualitatively validated SMARLA on three well-known case studies widely used in DRL research. Empirical results reveal that SMARLA is accurate at predicting safety violations, with a low false positive rate, and can predict violations at an early stage, approximately halfway through the execution of the agent, before violations occur. We also discuss different decision criteria, based on confidence intervals of the predicted violation probabilities, to trigger safety mechanisms aiming at a trade-off between early detection and low false positive rates.

Successful deployment of Deep Neural Networks (DNNs) requires their validation with an adequate test set to ensure a sufficient degree of confidence in test outcomes. Although well-established test adequacy assessment techniques have been proposed for DNNs, we still need to investigate their application within a comprehensive methodology for accurately predicting the fault detection ability of test sets and thus assessing their adequacy. In this paper, we propose and evaluate TEASMA, a comprehensive and practical methodology designed to accurately assess the adequacy of test sets for DNNs. In practice, TEASMA allows engineers to decide whether they can trust high-accuracy test results and thus validate the DNN before its deployment. Based on a DNN model's training set, TEASMA provides a procedure to build accurate DNN-specific prediction models of the Fault Detection Rate (FDR) of a test set using an existing adequacy metric, thus enabling its assessment. We evaluated TEASMA with four state-of-the-art test adequacy metrics: Distance-based Surprise Coverage (DSC), Likelihood-based Surprise Coverage (LSC), Input Distribution Coverage (IDC), and Mutation Score (MS). Our extensive empirical evaluation across multiple DNN models and input sets such as ImageNet, reveals a strong linear correlation between the predicted and actual FDR values derived from MS, DSC, and IDC, with minimum R^2 values of 0.94 for MS and 0.90 for DSC and IDC. Furthermore, a low average Root Mean Square Error (RMSE) of 9% between actual and predicted FDR values across all subjects, when relying on regression analysis and MS, demonstrates the latter's superior accuracy when compared to DSC and IDC, with RMSE values of 0.17 and 0.18, respectively. Overall, these results suggest that TEASMA provides a reliable basis for confidently deciding whether to trust test results for DNN models.

Growth in system complexity increases the need for automated techniques dedicated to different log analysis tasks such as Log-based Anomaly Detection (LAD). The latter has been widely addressed in the literature, mostly by means of a variety of deep learning techniques. Despite their many advantages, that focus on deep learning techniques is somewhat arbitrary as traditional Machine Learning (ML) techniques may perform well in many cases, depending on the context and datasets. In the same vein, semi-supervised techniques deserve the same attention as supervised techniques since the former have clear practical advantages. Further, current evaluations mostly rely on the assessment of detection accuracy. However, this is not enough to decide whether or not a specific ML technique is suitable to address the LAD problem in a given context. Other aspects to consider include training and prediction times as well as the sensitivity to hyperparameter tuning, which in practice matters to engineers. In this paper, we present a comprehensive empirical study, in which we evaluate supervised and semi-supervised, traditional and deep ML techniques w.r.t. four evaluation criteria: detection accuracy, time performance, sensitivity of detection accuracy and time performance to hyperparameter tuning. The experimental results show that supervised traditional and deep ML techniques fare similarly in terms of their detection accuracy and prediction time. Moreover, overall, sensitivity analysis to hyperparameter tuning w.r.t. detection accuracy shows that supervised traditional ML techniques are less sensitive than deep learning techniques. Further, semi-supervised techniques yield significantly worse detection accuracy than supervised techniques.

Automated logging statement generation supports developers in documenting critical software runtime behavior. Given the great success in natural language generation and programming language comprehension, large language models (LLMs) might help developers generate logging statements, but this has not yet been investigated. To fill the gap, this paper performs the first study on exploring LLMs for logging statement generation.We first build a logging statement generation dataset, LogBench, with two parts: (1) LogBench-O: logging statements collected from GitHub repositories, and (2) LogBench-T: the transformed unseen code from LogBench-O. Then, we leverage LogBench to evaluate the effectiveness and generalization capabilities (using LogBench-T) of eleven top-performing LLMs. In addition, we examine the performance of these LLMs against classical retrieval-based and machine learning-based logging methods from the era preceding LLMs. We further evaluate LLM's logging generalization capabilities using unseen data (LogBench-T) derived from code transformation techniques. While existing LLMs deliver decent predictions on logging levels and logging variables, our study indicates that they only achieve a maximum BLEU score of 0.249, thus calling for improvements. The paper also highlights the importance of prompt constructions and external factors (e.g., programming contexts and code comments) for LLMs' logging performance. Based on these findings, we identify five implications and provide practical advice for future logging research. Our empirical analysis discloses the limitations of current logging approaches while showcasing the potential of LLM-based logging tools, and provides actionable guidance for building more practical models.

Magnetic resonance spectroscopic imaging (MRSI) enables the simultaneous non-invasive acquisition of MR spectra from multiple spatial locations inside the brain. While 1H-MRSI is increasingly used in the human brain, it is not yet widely applied in the preclinical settings, mostly because of difficulties specifically related to very small nominal voxel size in the rodent brain and low concentration of brain metabolites, resulting in low signal-to-noise ratio SNR. In this context, we implemented a free induction decay 1H-MRSI sequence (1H-FID-MRSI) in the rat brain at 14.1T. We combined the advantages of 1H-FID-MRSI with the ultra-high magnetic field to achieve higher SNR, coverage and spatial resolution in the rodent brain, and developed a custom dedicated processing pipeline with a graphical user interface: MRS4Brain toolbox. LCModel fit, using the simulated metabolite basis-set and in-vivo measured MM, provided reliable fits for the data at acquisition delays of 1.3 and 0.94 ms. The resulting Cram\'er-Rao lower bounds were sufficiently low (<30%) for eight metabolites of interest, leading to highly reproducible metabolic maps. Similar spectral quality and metabolic maps were obtained between 1 and 2 averages, with slightly better contrast and brain coverage due to increased SNR in the latter case. Furthermore, the obtained metabolic maps were accurate enough to confirm the previously known brain regional distribution of some metabolites. The acquisitions proved high reproducibility over time. We demonstrated that the increased SNR and spectral resolution at 14.1T can be translated into high spatial resolution in 1H-FID-MRSI of the rat brain in 13 minutes, using the sequence and processing pipeline described herein. High-resolution 1H-FID-MRSI at 14.1T provided reproducible and high-quality metabolic mapping of brain metabolites with significantly reduced technical limitations., Comment: Dunja Simicic and Brayan Alves are joint first authors

Flaky tests are problematic because they non-deterministically pass or fail for the same software version under test, causing confusion and wasting development effort. While machine learning models have been used to predict flakiness and its root causes, there is much less work on providing support to fix the problem. To address this gap, in this paper, we focus on predicting the type of fix that is required to remove flakiness and then repair the test code on that basis. We do this for a subset of flaky tests where the root cause of flakiness is in the test itself and not in the production code. One key idea is to guide the repair process with additional knowledge about the test's flakiness in the form of its predicted fix category. Thus, we first propose a framework that automatically generates labeled datasets for 13 fix categories and trains models to predict the fix category of a flaky test by analyzing the test code only. Our experimental results using code models and few-shot learning show that we can correctly predict most of the fix categories. To show the usefulness of such fix category labels for automatically repairing flakiness, we augment the prompts of GPT-3.5 Turbo, a Large Language Model (LLM), with such extra knowledge to request repair suggestions. The results show that our suggested fix category labels, complemented with in-context learning, significantly enhance the capability of GPT-3.5 Turbo in generating fixes for flaky tests. Based on the execution and analysis of a sample of GPT-repaired flaky tests, we estimate that a large percentage of such repairs (roughly between 51% and 83%) can be expected to pass. For the failing repaired tests, on average, 16% of the test code needs to be further changed for them to pass., Comment: 26 pages, 20 Figures

Software systems log massive amounts of data, recording important runtime information. Such logs are used, for example, for log-based anomaly detection, which aims to automatically detect abnormal behaviors of the system under analysis by processing the information recorded in its logs. Many log-based anomaly detection techniques based on deep learning models include a pre-processing step called log parsing. However, understanding the impact of log parsing on the accuracy of anomaly detection techniques has received surprisingly little attention so far. Investigating what are the key properties log parsing techniques should ideally have to help anomaly detection is therefore warranted. In this paper, we report on a comprehensive empirical study on the impact of log parsing on anomaly detection accuracy, using 13 log parsing techniques, seven anomaly detection techniques (five based on deep learning and two based on traditional machine learning) on three publicly available log datasets. Our empirical results show that, despite what is widely assumed, there is no strong correlation between log parsing accuracy and anomaly detection accuracy, regardless of the metric used for measuring log parsing accuracy. Moreover, we experimentally confirm existing theoretical results showing that it is a property that we refer to as distinguishability in log parsing results as opposed to their accuracy that plays an essential role in achieving accurate anomaly detection.

The $\mathit{SU}(3)$ tensor multiplicities are piecewise polynomial of degree $1$ in their labels. The pieces are the chambers of a complex of cones. We describe in detail this chamber complex and determine the group of all linear symmetries (of order $144$) for these tensor multiplicities. We represent the cells by diagrams showing clearly the inclusions as well as the actions of the group of symmetries and of its remarkable subgroups., Comment: One example added

Requirement specifications are typically written in natural language (NL) due to its usability across multiple domains and understandability by all stakeholders. However, unstructured NL is prone to quality problems (e.g., ambiguity) in writing requirements, which can result in project failures. To address this issue, we present a tool, named Paska, that automatically detects quality problems as smells in NL requirements and offers recommendations to improve their quality. Our approach relies on natural language processing (NLP) techniques and, most importantly, a state-of-the-art controlled natural language (CNL) for requirements (Rimay), to detect smells and suggest recommendations using patterns defined in Rimay to improve requirement quality. We evaluated Paska through an industrial case study in the financial domain involving 13 systems and 2725 annotated requirements. The results show that our tool is accurate in detecting smells (precision of 89% and recall of 89%) and suggesting appropriate Rimay pattern recommendations (precision of 96% and recall of 94%).

Increased capabilities such as recognition and self-adaptability are now required from IoT applications. While IoT node power consumption is a major concern for these applications, cloud-based processing is becoming unsustainable due to continuous sensor or image data transmission over the wireless network. Thus optimized ML capabilities and data transfers should be integrated in the IoT node. Moreover, IoT applications are torn between sporadic data-logging and energy-hungry data processing (e.g. image classification). Thus, the versatility of the node is key in addressing this wide diversity of energy and processing needs. This paper presents SamurAI, a versatile IoT node bridging this gap in processing and in energy by leveraging two on-chip sub-systems: a low power, clock-less, event-driven Always-Responsive (AR) part and an energy-efficient On-Demand (OD) part. AR contains a 1.7MOPS event-driven, asynchronous Wake-up Controller (WuC) with a 207ns wake-up time optimized for sporadic computing, while OD combines a deep-sleep RISC-V CPU and 1.3TOPS/W Machine Learning (ML) for more complex tasks up to 36GOPS. This architecture partitioning achieves best in class versatility metrics such as peak performance to idle power ratio. On an applicative classification scenario, it demonstrates system power gains, up to 3.5x compared to cloud-based processing, and thus extended battery lifetime.

Test suites tend to grow when software evolves, making it often infeasible to execute all test cases with the allocated testing budgets, especially for large software systems. Test suite minimization (TSM) is employed to improve the efficiency of software testing by removing redundant test cases, thus reducing testing time and resources, while maintaining the fault detection capability of the test suite. Most existing TSM approaches rely on code coverage (white-box) or model-based features, which are not always available to test engineers. Recent TSM approaches that rely only on test code (black-box) have been proposed, such as ATM and FAST-R. To address the scalability, we propose LTM (Language model-based Test suite Minimization), a novel, scalable, and black-box similarity-based TSM approach based on large language models (LLMs), which is the first application of LLMs in the context of TSM. To support similarity measurement for test code embeddings, we investigate five pre-trained language models: CodeBERT, GraphCodeBERT, UniXcoder, StarEncoder, and CodeLlama, on which we compute two similarity measures: Cosine Similarity and Euclidean Distance. Our goal is to find similarity measures that are not only computationally more efficient but can also better guide a Genetic Algorithm (GA) to search for optimal minimized test suites, thus reducing the overall search time. Experimental results show that the best configuration of LTM (UniXcoder/Cosine) outperforms ATM in three aspects: (a) achieving a slightly greater saving rate of testing time (41.72% versus 41.02%, on average); (b) attaining a significantly higher fault detection rate (0.84 versus 0.81, on average); and, most importantly, (c) minimizing test suites nearly five times faster on average, with higher gains for larger test suites and systems, thus achieving much higher scalability.