1. A cross-benchmark examination of feature-based algorithm selector generalization in single-objective numerical optimizationGjorgjina Cenikj, Gašper Petelin, Tome Eftimov, 2024, original scientific article Abstract: The task of selecting the best optimization algorithm for a particular problem is known as algorithm selection (AS). This involves training a model using landscape characteristics to predict algorithm performance, but a key challenge remains: making AS models generalize effectively to new, untrained benchmark suites. This study assesses AS models’ generalizability in single-objective numerical optimization across diverse benchmark suites. Using Exploratory Landscape Analysis (ELA) and transformer-based (TransOpt) features, the research investigates their individual and combined effectiveness in AS across four benchmarks: BBOB, AFFINE, RANDOM, and ZIGZAG. AS models perform differently based on benchmark suite similarities in algorithm performance distributions and single-best solvers. When suites align, these models underperform against a baseline predicting mean algorithm performance; yet, they outperform the baseline when suites differ in performance distributions and solvers. The AS models trained using the ELA landscape features are better than the models trained using the TransOpt features on the BBOB and AFFINE benchmark suites, while the opposite is true for the RANDOM benchmark suite. Ultimately, the study reveals challenges in accurately capturing algorithm performance through problem landscape features (ELA or TransOpt), impacting AS model applicability.
Keywords: algorithm selection, multi-target regression, generalization, benchmarking
Published in DiRROS: 21.05.2024; Views: 41; Downloads: 140
 Full text (2,49 MB)
This document has many files! More... |
2. Models for forecasting the traffic flow within the city of LjubljanaGašper Petelin, Rok Hribar, Gregor Papa, 2023, original scientific article Abstract: Efficient traffic management is essential in modern urban areas. The development of intelligent traffic flow prediction systems can help to reduce travel times and maximize road capacity utilization. However, accurately modeling complex spatiotemporal dependencies can be a difficult task, especially when real-time data collection is not possible. This study aims to tackle this challenge by proposing a solution that incorporates extensive feature engineering to combine historical traffic patterns with covariates such as weather data and public holidays. The proposed approach is assessed using a new real-world data set of traffic patterns collected in Ljubljana, Slovenia. The constructed models are evaluated for their accuracy and hyperparameter sensitivity, providing insights into their performance. By providing practical solutions for real-world scenarios, the proposed approach offers an effective means to improve traffic flow prediction without relying on real-time data.
Keywords: traffic modeling, time-series forecasting, traffic-count data set, machine learning, model comparison
Published in DiRROS: 28.09.2023; Views: 376; Downloads: 158
Full text (5,05 MB)
This document has many files! More... |
3. DynamoRep : trajectory-based population dynamics for classification of black-box optimization problemsGjorgjina Cenikj, Gašper Petelin, Carola Doerr, Peter Korošec, Tome Eftimov, 2023, published scientific conference contribution Keywords: black-box single-objective optimization, optimization problem classification, problem representation, meta-learning
Published in DiRROS: 30.08.2023; Views: 346; Downloads: 236
Full text (650,13 KB)
This document has many files! More... |
4. |
5. |
6. |
7. |
8. |
9. |
