arXiv subjects

Matthias Weidlich

Publications and source records attributed to Matthias Weidlich.

At least 19 recordsLinked to original sources

arXiv2026

FINER-SQL: Boosting Small Language Models for Text-to-SQL

Large language models have driven major advances in Text-to-SQL generation. However, they suffer from high computational cost, long latency, and data privacy concerns, which make them impractical for many real-world applications. A natural alternative is to use small language models (SLMs), which enable efficient and private on-premise deployment. Yet, SLMs often struggle with weak reasoning and poor instruction following. Conventional reinforcement learning methods based on sparse binary rewards (0/1) provide little learning signal when the generated SQLs are incorrect, leading to unstable or collapsed training. To overcome these issues, we propose FINER-SQL, a scalable and reusable reinforcement learning framework that enhances SLMs through fine-grained execution feedback. Built on group relative policy optimization, FINER-SQL replaces sparse supervision with dense and interpretable rewards that offer continuous feedback even for incorrect SQLs. It introduces two key reward functions: a memory reward, which aligns reasoning with verified traces for semantic stability, and an atomic reward, which measures operation-level overlap to grant partial credit for structurally correct but incomplete SQLs. This approach transforms discrete correctness into continuous learning, enabling stable, critic-free optimization. Experiments on the BIRD and Spider benchmarks show that FINER-SQL achieves up to 67.73\% and 85\% execution accuracy with a 3B model -- matching much larger LLMs while reducing inference latency to 5.57~s/sample. These results highlight a cost-efficient and privacy-preserving path toward high-performance Text-to-SQL generation. Our code is available at https://github.com/thanhdath/finer-sql.

cs.DB↗

arXiv2025

A Multi-agent Text2SQL Framework using Small Language Models and Execution Feedback

Text2SQL, the task of generating SQL queries from natural language text, is a critical challenge in data engineering. Recently, Large Language Models (LLMs) have demonstrated superior performance for this task due to their advanced comprehension and generation capabilities. However, privacy and cost considerations prevent companies from using Text2SQL solutions based on external LLMs offered as a service. Rather, small LLMs (SLMs) that are openly available and can hosted in-house are adopted. These SLMs, in turn, lack the generalization capabilities of larger LLMs, which impairs their effectiveness for complex tasks such as Text2SQL. To address these limitations, we propose MATS, a novel Text2SQL framework designed specifically for SLMs. MATS uses a multi-agent mechanism that assigns specialized roles to auxiliary agents, reducing individual workloads and fostering interaction. A training scheme based on reinforcement learning aligns these agents using feedback obtained during execution, thereby maintaining competitive performance despite a limited LLM size. Evaluation results using on benchmark datasets show that MATS, deployed on a single- GPU server, yields accuracy that are on-par with large-scale LLMs when using significantly fewer parameters. Our source code and data are available at https://github.com/thanhdath/mats-sql.

cs.DB↗

arXiv2025

ContinuumConductor : Decentralized Process Mining on the Edge-Cloud Continuum

Process mining traditionally assumes centralized event data collection and analysis. However, modern Industrial Internet of Things systems increasingly operate over distributed, resource-constrained edge-cloud infrastructures. This paper proposes a structured approach for decentralizing process mining by enabling event data to be mined directly within the IoT systems edge-cloud continuum. We introduce ContinuumConductor a layered decision framework that guides when to perform process mining tasks such as preprocessing, correlation, and discovery centrally or decentrally. Thus, enabling privacy, responsive and resource-efficient process mining. For each step in the process mining pipeline, we analyze the trade-offs of decentralization versus centralization across these layers and propose decision criteria. We demonstrate ContinuumConductor at a real-world use-case of process optimazition in inland ports. Our contributions lay the foundation for computing-aware process mining in cyber-physical and IIoT systems.

cs.DC↗

arXiv2025

Determining Window Sizes using Species Estimation for Accurate Process Mining over Streams

Streaming process mining deals with the real-time analysis of event streams. A common approach for it is to adopt windowing mechanisms that select event data from a stream for subsequent analysis. However, the size of these windows denotes a crucial parameter, as it influences the representativeness of the window content and, by extension, of the analysis results. Given that process dynamics are subject to changes and potential concept drift, a static, fixed window size leads to inaccurate representations that introduce bias in the analysis. In this work, we present a novel approach for streaming process mining that addresses these limitations by adjusting window sizes. Specifically, we dynamically determine suitable window sizes based on estimators for the representativeness of samples as developed for species estimation in biodiversity research. Evaluation results on real-world data sets show improvements over existing approaches that adopt static window sizes in terms of accuracy and robustness to concept drifts.

cs.DB↗

arXiv2025

The Past Still Matters: A Temporally-Valid Data Discovery System

Over the past decade, the proliferation of public and enterprise data lakes has fueled intensive research into data discovery, aiming to identify the most relevant data from vast and complex corpora to support diverse user tasks. Significant progress has been made through the development of innovative index structures, similarity measures, and querying infrastructures. Despite these advances, a critical aspect remains overlooked: relevance is time-varying. Existing discovery methods largely ignore this temporal dimension, especially when explicit date/time metadata is missing. To fill this gap, we outline a vision for a data discovery system that incorporates the temporal dimension of data. Specifically, we define the problem of temporally-valid data discovery and argue that addressing it requires techniques for version discovery, temporal lineage inference, change log synthesis, and time-aware data discovery. We then present a system architecture to deliver these techniques, before we summarize research challenges and opportunities. As such, we lay the foundation for a new class of data discovery systems, transforming how we interact with evolving data lakes.

cs.DB↗

arXiv2025

Bridging Imperative Process Models and Process Data Queries-Translation and Relaxation

Business process management is increasingly practiced using data-driven approaches. Still, classical imperative process models, which are typically formalized using Petri nets, are not straightforwardly applicable to the relational databases that contain much of the available structured process execution data. This creates a gap between the traditional world of process modeling and recent developments around data-driven process analysis, ultimately leading to the under-utilization of often readily available process models. In this paper, we close this gap by providing an approach for translating imperative models into relaxed process data queries, specifically SQL queries executable on relational databases, for conformance checking. Our results show the continued relevance of imperative process models to data-driven process management, as well as the importance of behavioral footprints and other declarative approaches for integrating model-based and data-driven process management.

cs.DB↗

arXiv2025

Convolution and Graph-based Deep Learning Approaches for Gamma/Hadron Separation in Imaging Atmospheric Cherenkov Telescopes

The identification of $\gamma$-rays from the predominant hadronic-background is a key aspect in their ground-based detection using Imaging Atmospheric Cherenkov Telescopes (IACTs). While current methods are limited in their ability to exploit correlations in complex data, deep learning-based models offer a promising alternative by directly leveraging image-level information. However, several challenges involving the robustness and applicability of such models remain. Designing model architectures with inductive biases relevant for the task can help mitigate the problem. Three such deep learning-based models are proposed, trained, and evaluated on simulated data: (1) a hybrid convolutional and graph neural network model (CNN-GNN) using both image and graph data; (2) an enhanced CNN-GNN variant that incorporates additional reconstructed information within the graph construction; and (3) a graph neural network (GNN) model using image moments serving as a baseline. The new combined convolution and graph-based approach demonstrates improved performance over traditional methods, and the inclusion of reconstructed information offers further potential in generalization capabilities on real observational data.

astro-ph.HE↗

arXiv2025

Let's Simply Count: Quantifying Distributional Similarity Between Activities in Event Data

To obtain insights from event data, advanced process mining methods assess the similarity of activities to incorporate their semantic relations into the analysis. Here, distributional similarity that captures similarity from activity co-occurrences is commonly employed. However, existing work for distributional similarity in process mining adopt neural network-based approaches as developed for natural language processing, e.g., word2vec and autoencoders. While these approaches have been shown to be effective, their downsides are high computational costs and limited interpretability of the learned representations. In this work, we argue for simplicity in the modeling of distributional similarity of activities. We introduce count-based embeddings that avoid a complex training process and offer a direct interpretable representation. To underpin our call for simple embeddings, we contribute a comprehensive benchmarking framework, which includes means to assess the intrinsic quality of embeddings, their performance in downstream applications, and their computational efficiency. In experiments that compare against the state of the art, we demonstrate that count-based embeddings provide a highly effective and efficient basis for distributional similarity between activities in event data.

cs.DB↗

arXiv2025

SHARP: Shared State Reduction for Efficient Matching of Sequential Patterns

The detection of sequential patterns in data is a basic functionality of modern data processing systems for complex event processing (CEP), OLAP, and retrieval-augmented generation (RAG). In practice, pattern matching is challenging, since common applications rely on a large set of patterns that shall be evaluated with tight latency bounds. At the same time, matching needs to maintain state, i.e., intermediate results, that grows exponentially in the input size. Hence, systems turn to best-effort processing, striving for maximal recall under a latency bound. Existing techniques, however, consider each pattern in isolation, neglecting the optimization potential induced by state sharing in pattern matching. In this paper, we present SHARP, a library that employs state reduction to achieve efficient best-effort pattern matching. To this end, SHARP incorporates state sharing between patterns through a new abstraction, coined pattern-sharing degree (PSD). At runtime, this abstraction facilitates the categorization and indexing of partial pattern matches. Based thereon, once a latency bound is exceeded, SHARP realizes best-effort processing by selecting a subset of partial matches for further processing in constant time. In experiments with real-world data, SHARP achieves a recall of 97%, 96% and 73% for pattern matching in CEP, OLAP, and RAG applications, under a bound of 50% of the average processing latency.

cs.DB↗

arXiv2025

Process Mining on Distributed Data Sources

Major domains such as logistics, healthcare, and smart cities increasingly rely on sensor technologies and distributed infrastructures to monitor complex processes in real time. These developments are transforming the data landscape from discrete, structured records stored in centralized systems to continuous, fine-grained, and heterogeneous event streams collected across distributed environments. As a result, traditional process mining techniques, which assume centralized event logs from enterprise systems, are no longer sufficient. In this paper, we discuss the conceptual and methodological foundations for this emerging field. We identify three key shifts: from offline to online analysis, from centralized to distributed computing, and from event logs to sensor data. These shifts challenge traditional assumptions about process data and call for new approaches that integrate infrastructure, data, and user perspectives. To this end, we define a research agenda that addresses six interconnected fields, each spanning multiple system dimensions. We advocate a principled methodology grounded in algorithm engineering, combining formal modeling with empirical evaluation. This approach enables the development of scalable, privacy-aware, and user-centric process mining techniques suitable for distributed environments. Our synthesis provides a roadmap for advancing process mining beyond its classical setting, toward a more responsive and decentralized paradigm of process intelligence.

cs.ET↗

arXiv2024

Early Detection of Multiwavelength Blazar Variability

Blazars are a subclass of active galactic nuclei with relativistic jets pointing toward the observer. They are notable for their flux variability at all observed wavelengths and timescales. Together with simultaneous measurements at lower energies, the very-high-energy (VHE) emission observed during blazar flares may be used to probe the population of accelerated particles. However, optimally triggering observations of blazar high states can be challenging. Notable examples include identifying a flaring episode in real time and predicting VHE flaring activity based on lower-energy observables. For this purpose, we have developed a novel deep learning analysis framework, based on data-driven anomaly detection techniques. It is capable of detecting various types of anomalies in real-world, multiwavelength light curves, ranging from clear high states to subtle correlations across bands. Based on unsupervised anomaly detection and clustering methods, we differentiate source variability from noisy background activity, without the need for a labeled training data set of flaring states. The framework incorporates measurement uncertainties and is robust given data quality challenges, such as varying cadences and observational gaps. We evaluate our approach using both historical data and simulations of blazar light curves in two energy bands, corresponding to sources observable with the Fermi Large Area Telescope and the upcoming Cherenkov Telescope Array Observatory. In a statistical analysis, we show that our framework can reliably detect known historical flares.

astro-ph.HE↗

arXiv2024

Control-flow Reconstruction Attacks on Business Process Models

Process models may be automatically generated from event logs that contain as-is data of a business process. While such models generalize over the control-flow of specific, recorded process executions, they are often also annotated with behavioural statistics, such as execution frequencies.Based thereon, once a model is published, certain insights about the original process executions may be reconstructed, so that an external party may extract confidential information about the business process. This work is the first to empirically investigate such reconstruction attempts based on process models. To this end, we propose different play-out strategies that reconstruct the control-flow from process trees, potentially exploiting frequency annotations. To assess the potential success of such reconstruction attacks on process models, and hence the risks imposed by publishing them, we compare the reconstructed process executions with those of the original log for several real-world datasets.

cs.DB↗

arXiv2024

Fast-FedUL: A Training-Free Federated Unlearning with Provable Skew Resilience

Federated learning (FL) has recently emerged as a compelling machine learning paradigm, prioritizing the protection of privacy for training data. The increasing demand to address issues such as ``the right to be forgotten'' and combat data poisoning attacks highlights the importance of techniques, known as \textit{unlearning}, which facilitate the removal of specific training data from trained FL models. Despite numerous unlearning methods proposed for centralized learning, they often prove inapplicable to FL due to fundamental differences in the operation of the two learning paradigms. Consequently, unlearning in FL remains in its early stages, presenting several challenges. Many existing unlearning solutions in FL require a costly retraining process, which can be burdensome for clients. Moreover, these methods are primarily validated through experiments, lacking theoretical assurances. In this study, we introduce Fast-FedUL, a tailored unlearning method for FL, which eliminates the need for retraining entirely. Through meticulous analysis of the target client's influence on the global model in each round, we develop an algorithm to systematically remove the impact of the target client from the trained model. In addition to presenting empirical findings, we offer a theoretical analysis delineating the upper bound of our unlearned model and the exact retrained model (the one obtained through retraining using untargeted clients). Experimental results with backdoor attack scenarios indicate that Fast-FedUL effectively removes almost all traces of the target client, while retaining the knowledge of untargeted clients (obtaining a high accuracy of up to 98\% on the main task). Significantly, Fast-FedUL attains the lowest time complexity, providing a speed that is 1000 times faster than retraining. Our source code is publicly available at \url{https://github.com/thanhtrunghuynh93/fastFedUL}.

cs.LG↗

arXiv2024

Manipulating Recommender Systems: A Survey of Poisoning Attacks and Countermeasures

Recommender systems have become an integral part of online services to help users locate specific information in a sea of data. However, existing studies show that some recommender systems are vulnerable to poisoning attacks, particularly those that involve learning schemes. A poisoning attack is where an adversary injects carefully crafted data into the process of training a model, with the goal of manipulating the system's final recommendations. Based on recent advancements in artificial intelligence, such attacks have gained importance recently. While numerous countermeasures to poisoning attacks have been developed, they have not yet been systematically linked to the properties of the attacks. Consequently, assessing the respective risks and potential success of mitigation strategies is difficult, if not impossible. This survey aims to fill this gap by primarily focusing on poisoning attacks and their countermeasures. This is in contrast to prior surveys that mainly focus on attacks and their detection methods. Through an exhaustive literature review, we provide a novel taxonomy for poisoning attacks, formalise its dimensions, and accordingly organise 30+ attacks described in the literature. Further, we review 40+ countermeasures to detect and/or prevent poisoning attacks, evaluating their effectiveness against specific types of attacks. This comprehensive survey should serve as a point of reference for protecting recommender systems against poisoning attacks. The article concludes with a discussion on open issues in the field and impactful directions for future research. A rich repository of resources associated with poisoning attacks is available at https://github.com/tamlhp/awesome-recsys-poisoning.

cs.CR↗

arXiv2024

Mining a Minimal Set of Behavioral Patterns using Incremental Evaluation

Process mining provides methods to analyse event logs generated by information systems during the execution of processes. It thereby supports the design, validation, and execution of processes in domains ranging from healthcare, through manufacturing, to e-commerce. To explore the regularities of flexible processes that show a large behavioral variability, it was suggested to mine recurrent behavioral patterns that jointly describe the underlying process. Existing approaches to behavioral pattern mining, however, suffer from two limitations. First, they show limited scalability as incremental computation is incorporated only in the generation of pattern candidates, but not in the evaluation of their quality. Second, process analysis based on mined patterns shows limited effectiveness due to an overwhelmingly large number of patterns obtained in practical application scenarios, many of which are redundant. In this paper, we address these limitations to facilitate the analysis of complex, flexible processes based on behavioral patterns. Specifically, we improve COBPAM, our initial behavioral pattern mining algorithm, by an incremental procedure to evaluate the quality of pattern candidates, optimizing thereby its efficiency. Targeting a more effective use of the resulting patterns, we further propose pruning strategies for redundant patterns and show how relations between the remaining patterns are extracted and visualized to provide process insights. Our experiments with diverse real-world datasets indicate a considerable reduction of the runtime needed for pattern mining, while a qualitative assessment highlights how relations between patterns guide the analysis of the underlying process.

cs.DB↗

arXiv2023

Knowledge-Driven Modulation of Neural Networks with Attention Mechanism for Next Activity Prediction

Predictive Process Monitoring (PPM) aims at leveraging historic process execution data to predict how ongoing executions will continue up to their completion. In recent years, PPM techniques for the prediction of the next activities have matured significantly, mainly thanks to the use of Neural Networks (NNs) as a predictor. While their performance is difficult to beat in the general case, there are specific situations where background process knowledge can be helpful. Such knowledge can be leveraged for improving the quality of predictions for exceptional process executions or when the process changes due to a concept drift. In this paper, we present a Symbolic[Neuro] system that leverages background knowledge expressed in terms of a procedural process model to offset the under-sampling in the training data. More specifically, we make predictions using NNs with attention mechanism, an emerging technology in the NN field. The system has been tested on several real-life logs showing an improvement in the performance of the prediction task.

cs.AI↗

arXiv2023

How do users design scientific workflows? The Case of Snakemake

Scientific workflows automate the analysis of large-scale scientific data, fostering the reuse of data processing operators as well as the reproducibility and traceability of analysis results. In exploratory research, however, workflows are continuously adapted, utilizing a wide range of tools and software libraries, to test scientific hypotheses. Script-based workflow engines cater to the required flexibility through direct integration of programming primitives but lack abstractions for interactive exploration of the workflow design by a user during workflow execution. To derive requirements for such interactive workflows, we conduct an empirical study on the use of Snakemake, a popular Python-based workflow engine. Based on workflows collected from 1602 GitHub repositories, we present insights on common structures of Snakemake workflows, as well as the language features typically adopted in their specification.

cs.OH↗

arXiv2023

Large Process Models: A Vision for Business Process Management in the Age of Generative AI

The continued success of Large Language Models (LLMs) and other generative artificial intelligence approaches highlights the advantages that large information corpora can have over rigidly defined symbolic models, but also serves as a proof-point of the challenges that purely statistics-based approaches have in terms of safety and trustworthiness. As a framework for contextualizing the potential, as well as the limitations of LLMs and other foundation model-based technologies, we propose the concept of a Large Process Model (LPM) that combines the correlation power of LLMs with the analytical precision and reliability of knowledge-based systems and automated reasoning approaches. LPMs are envisioned to directly utilize the wealth of process management experience that experts have accumulated, as well as process performance data of organizations with diverse characteristics, e.g.,\ regarding size, region, or industry. In this vision, the proposed LPM would allow organizations to receive context-specific (tailored) process and other business models, analytical deep-dives, and improvement recommendations. As such, they would allow to substantially decrease the time and effort required for business transformation, while also allowing for deeper, more impactful, and more actionable insights than previously possible. We argue that implementing an LPM is feasible, but also highlight limitations and research challenges that need to be solved to implement particular aspects of the LPM vision.

cs.SE↗