14 results
Now showing1 - 10 of 14
- Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: with files , Can explainability metrics improve genetic programming? Lessons from 2048(Association for Computing Machinery (ACM), 2026-08-13); ; ; Genetic programming (GP) produces solutions that, while often less performant than neural networks, are uniquely amenable to analysis. In domains like game-playing, this explainability could reveal why certain policies succeed or fail-but such insights are rarely leveraged to improve the algorithms themselves. Here, we analyze a GP algorithm for 2048, testing whether structural, behavioral, or semantic metrics of evolved policies correlate with performance. While structural metrics showed no predictive power and behavioral features yielded ambiguous results, we identified a semantic feature that correlated with policy quality. Using this insight, we designed a mutation operator that improved performance. Though modest, this improvement suggests that explainability metrics can guide operator design, even when broader explanatory goals remain unmet. More broadly, our work highlights a potential advantage of GP: its solutions may be analyzable in ways that opaque methods like neural networks are not.Publicationtype: Conference Paper2 3 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: with files , Ride-sharing simulation for optimisation(Association for Computing Machinery (ACM), 2026-08-13); ; Optimisation research requires possibilities to test and evaluate new algorithms and algorithm adaptions. Two interesting problem classes-relevant to society-that covers shortest path problems and resource allocation in a very dynamic environment, with the possibility of running endlessly, are ride-hailing and ride-sharing. This paper presents an extensible simulation framework for ride-hailing and ride-sharing problems that allows researchers to create solutions in any program language, provides a graphical user-interface, and automatic evaluation.Publicationtype: Conference Paper2 3 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: with files , Enhancing surrogate model usability for optimisation experts through extended ML support in EvoAl(Association for Computing Machinery (ACM), 2026-08-13); ; ; Surrogate models are a crucial aspect in many real-world application optimisation problems. Thus, optimisation experts do not only have to focus on the optimisation problem at hand, but also work on selecting and configuring appropriate models. This may be challenging, when no extensive expertise in machine learning (ML) is present. To address this issue, we extend the open-source data science research tool EvoAl to better support optimisation practitioners in focusing on their core tasks. Specifically, we enhance EvoAl’s machine learning language to accommodate the growing complexity of surrogate modelling workflows. Our contributions include expanding the range of available models, introducing runtime validation of model configurations, and enabling the integration of pre-trained models via ONNX. These improvements lower the barrier to applying advanced ML techniques in optimisation, thereby facilitating more efficient and robust surrogate-based optimisation in practical applications.Publicationtype: Conference Paper5 3 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , A comparison of repositioning and scheduling algorithms for the ride-hailing problem(Springer, 2026-08); ; ; The Ride-Hailing Problem is an online optimization challenge that coordinates a fleet of vehicles to serve requests revealed over time. The problem is NP-hard, and many solution approaches use stochastic or machine-learning based algorithms, which perform well but are difficult to interpret. We study the trade-offs between performance and explainability. Our results show that greedy assignment consistently outperforms more complex methods, indicating that explainability need not come at the cost of efficiency. We also find that the effectiveness of repositioning strategies strongly depends on demand patterns, highlighting the value of context-aware policies.Publicationtype: Conference Paper7 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Explanation in bio-inspired computing: towards understanding of AI systemsArtificial intelligence methods and applications have recently seen a massive surge, partially caused by the success of neural networks in areas like image classification and LLMs for generating near-perfect natural-language texts. Unnoticed by the public, but highly important for many AI methods to function, bio-inspired optimisation techniques have also seen a rising usage. However, the more techniques are used, the more complex the explainability decreases. Even developers of Neural Networks can seldom state why the Neural Network’s results are what it is. The explainability of AI methods, as well as systems in general, is, however, essential for safety, security reasons, and to gain and maintain trust with system users. While research in explainability has therefore gained significant traction with prominent AI methods, such as neural networks, bio-inspired optimisation techniques have seen less research in this regard. The complexity of explainability with these algorithms lies in the use of populations and randomness. We present an approach to track individuals in bio-inspired optimisation techniques, aiming to improve our understanding of the quality of results from such optimisation algorithms. To that end, we introduce a data model, include this model in the standard implementations of these approaches, and provide a visualisation that allows for understanding the relational information of these individuals, yielding more insight into these optimisation techniques and providing a first step toward improved explainability.Publicationtype: Conference Paper7 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Why less is sometimes more: using Boolean literals to solve 2048Explaining and understanding AI-generated policies of control problems is crucial for the acceptance of such policies. Based on an optimisation challenge that took place at GECCO 2024, we describe different solutions for optimising generated policies for the well-known game 2048. At the same time, these generated policies aim to be simpler to understand and, thus, their decisions explainable in contrast to current solutions for 2048, as, e.g., neural network-based models. Our approach uses only Boolean expressions in the policy, and the optimisation shows that such a policy has advantages compared to more complex policy variants. The optimisation generates better results in a shorter time than for non-Boolean expressions. Additionally, the Boolean policies are smaller in size and can be reduced even more when applying existing techniques for term rewriting and simplification. These simplifications, again, may aid in understanding the policy's decision.Publicationtype: Journal Article74 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Finding the perfect MRI sequence for your patient : towards an optimisation workflow for MRI-sequences(IEEE, 2024-08-08); ; Günther, MatthiasMagnetic Resonance Imaging (MRI) is an essential tool for medical diagnosis. At the same time, its usage requires profound expert knowledge to determine the ideal MR sequence and protocol to be run. Until now, the contrast and quality of the resulting image have relied mainly on the radiologist's expertise. When confronted with clinical requirements and patient information, the radiologist chooses suitable sequence protocols for the examination. We propose a workflow that supports medical personnel in finding the optimal sequence for a given diagnostic task. To that end, we combine evolutionary algorithms for the optimisation, machine learning techniques for training a surrogate optimisation function from simulated MRI data, and domain-specific languages to allow non-programmers to formulate their requirements and constraints semi-formally. In this paper, we focus on the efficient usage of real-world application-motivated adaptions of the used evolutionary algorithm and evaluate their effects on four real-life sequence examples. We show that it is essential to use an adaption for the surrogate model to obtain realistic solutions and use correlation information about the search space to stay in feasible areas of the search space and thus improve optimisation quality. These findings are a first step in automating the entire MRI-sequence optimisation flow, which is necessary to allow a more widespread usage of this essential medical diagnostic technique.Publicationtype: Conference Paper19 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , EvoAl - codeless domain-optimisation(Association for Computing Machinery (ACM), 2024-07-14); ; ; Applying optimisation techniques such as evolutionary computation to real-world tasks often requires significant adaptation. However, specific application domains do not typically demand major changes to existing optimisation methods. The decisive aspect is the inclusion of domain knowledge and configuration of established techniques to suit the problem. Separating the optimisation technique from the domain knowledge offers several advantages: First, it allows updating domain knowledge without necessitating reimplementation. Second, it improves identification and comparison of the optimisation methods employed. We present EvoAl, an open-source data-science research tool suite that focuses on optimisation research for real-world problems. EvoAl implements the separation of domain-knowledge and detaches implementation from configuration, facilitating optimisation with little programming effort, allowing direct comparability with other approaches (using EvoAl), and ensuring reproducibility. EvoAl also includes options for surrogate models, data models for complex search spaces, data validation, and benchmarking options for optimisation researchers.Publicationtype: Conference Paper15 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Hybrid PTX analysis for GPU accelerated CNN inferencing aiding computer architecture designGeneral-Purpose Computation on Graphics Processing Units (GPGPUs) are becoming crucial in accelerating computing capacity. Due to the massive parallelism capabilities of GPUs, they can achieve impressive speedups of up to 32 times compared to common CPUs. However, writing highly parallel code and utilizing a GPU is challenging for programmers. Developers are facing new challenges since GPUs handle threads and parallelism differently from CPUs. Academia and industry proposed several profilers to support developers in terms of code optimization. These profilers often require an actual device (e.g., GPU) and take a long time for the profiling process. We propose HyPA, a hybrid Parallel Thread Execution (PTX) Analyzer that inspects PTX code statically and dynamically. HyPA implements a partly functional emulator that executes instructions that rely on runtime dependencies to count the number of executed PTX instructions and divergent branches. HyPa executes compiled kernels-the programs that run on GPUs-generated by the CUDA compiler and supports the full PTX 7.7 specification. Our functional emulator allows significantly faster analysis of PTX code compared to standard profilers. In our evaluation, we quantify this increase in performance through benchmark runs. HyPA achieved speedups of up to 536% compared to the nvprof profiler. Moreover, our approach can gather performance metrics beyond static analysis (e.g., branch efficiency) by a faster execution time than by profiling the application on an actual device. Finally, we provide an open-source implementation of HyPA to help developers and system designers in further research and development.Publicationtype: Conference Paper12 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Repetitive Processes and Their Surrogate-Model Congruent Encoding for Evolutionary Algorithms - A Theoretic Proposal(2023-07); ; Evolutionary algorithms are a well-known optimisation technique. They can handle very different optimisation tasks and deal with distorted search spaces as well as non-differentiable optimisation functions. One crucial aspect in the design of evolutionary algorithms is the choice of encoding. Especially its interplay with the other components of the evolutionary algorithm is a relevant factor for the success of an evolutionary algorithm. While some encoding situations are relatively trivial, others pose a challenge. We focus on encoding repetitive processes, i.e. processes that consist of several variations of the same basic process (only with varied parameters). Our work proposes a possible technique that enables the validity of the encoded search space. We also provide adaptions to the standard operators of evolutionary algorithms to ensure they produce valid solutions. Furthermore, we show how this encoding technique is compatible with using a surrogate function for the fitness calculation and may reduce the necessary training data.Publicationtype: Conference Paper12