Robert Manthey
Research Subjects
Robert Manthey has a total of 19 co-authors. Here are the top co-authors by number of shared publications:
- Christian Roschke
University of Applied Sciences Mittweida
6 shared publications - Marc Ritter
University of Applied Sciences Mittweida
6 shared publications - Matthias Vodel
Hochschule Mittweida,Faculty Applied Computer Sciences Biosciences,Mittweida,Germany,09648
4 shared publications - Richard Vogel
Hochschule Mittweida,Faculty Applied Computer Sciences Biosciences,Mittweida,Germany,09648
4 shared publications - Matthias Baumgart
Hochschule Mittweida,Faculty Applied Computer Sciences Biosciences,Mittweida,Germany,09648
3 shared publications Rico Thomanek
Hochschule Mittweida
3 shared publications
Latest Publications
- Open Access
- 9 April 2024
- Published by MDPI AG in Machine Learning and Knowledge Extraction
- Vol. 6 (2) , 800-826
A Meta Algorithm for Interpretable Ensemble Learning: The League of Experts
Background. The importance of explainable artificial intelligence and machine learning (XAI/XML) is increasingly being recognized, aiming to understand how information contributes to decisions, the method’s bias, or sensitivity to data pathologies. Efforts are often directed to post hoc explanations of black box models. These approaches add additional sources for errors without resolving their shortcomings. Less effort is directed into the design of intrinsically interpretable approaches. Methods. We introduce an intrinsically interpretable methodology motivated by ensemble learning: the League of Experts (LoE) model. We establish the theoretical framework first and then deduce a modular meta algorithm. In our description, we focus primarily on classification problems. However, LoE applies equally to regression problems. Specific to classification problems, we employ classical decision trees as classifier ensembles as a particular instance. This choice facilitates the derivation of human-understandable decision rules for the underlying classification problem, which results in a derived rule learning system denoted as RuleLoE. Results. In addition to 12 KEEL classification datasets, we employ two standard datasets from particularly relevant domains—medicine and finance—to illustrate the LoE algorithm. The performance of LoE with respect to its accuracy and rule coverage is comparable to common state-of-the-art classification methods. Moreover, LoE delivers a clearly understandable set of decision rules with adjustable complexity, describing the classification problem. Conclusions. LoE is a reliable method for classification and regression problems with an accuracy that seems to be appropriate for situations in which underlying causalities are in the center of interest rather than just accurate predictions or classifications.
- 9 April 2024
- Published by MDPI AG in Machine Learning and Knowledge Extraction
- Vol. 6 (2) , 800-826
Open AccessA Meta Algorithm for Interpretable Ensemble Learning: The League of Experts
Background. The importance of explainable artificial intelligence and machine learning (XAI/XML) is increasingly being recognized, aiming to understand how information contributes to decisions, the method’s bias, or sensitivity to data pathologies. Efforts are often directed to post hoc explanations of black box models. These approaches add additional sources for errors without resolving their shortcomings. Less effort is directed into the design of intrinsically interpretable approaches. Methods. We introduce an intrinsically interpretable methodology motivated by ensemble learning: the League of Experts (LoE) model. We establish the theoretical framework first and then deduce a modular meta algorithm. In our description, we focus primarily on classification problems. However, LoE applies equally to regression problems. Specific to classification problems, we employ classical decision trees as classifier ensembles as a particular instance. This choice facilitates the derivation of human-understandable decision rules for the underlying classification problem, which results in a derived rule learning system denoted as RuleLoE. Results. In addition to 12 KEEL classification datasets, we employ two standard datasets from particularly relevant domains—medicine and finance—to illustrate the LoE algorithm. The performance of LoE with respect to its accuracy and rule coverage is comparable to common state-of-the-art classification methods. Moreover, LoE delivers a clearly understandable set of decision rules with adjustable complexity, describing the classification problem. Conclusions. LoE is a reliable method for classification and regression problems with an accuracy that seems to be appropriate for situations in which underlying causalities are in the center of interest rather than just accurate predictions or classifications.
- Open Access
- 5 July 2023
- Published by Association for Computing Machinery (ACM)
- p. 120-122
Decentralized Resilient Smart Lock System with Offline Capabilities - ChainLock
The management of authorizations for access to rooms, buildings and similar is of great importance for many facilities and institutions. However, similar with increasing size and variability, the complexity and the effort required for updating also increases significantly. The presented decentralized solution with locks featuring no permanent connection represents a significant improvement and simplification in this regard. In addition, its distributed blockchain architecture significantly reduces the possibilities for manipulation.
- 5 July 2023
- Published by Association for Computing Machinery (ACM)
- p. 120-122
Open AccessDecentralized Resilient Smart Lock System with Offline Capabilities - ChainLock
The management of authorizations for access to rooms, buildings and similar is of great importance for many facilities and institutions. However, similar with increasing size and variability, the complexity and the effort required for updating also increases significantly. The presented decentralized solution with locks featuring no permanent connection represents a significant improvement and simplification in this regard. In addition, its distributed blockchain architecture significantly reduces the possibilities for manipulation.
- 20 July 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 19, 1-9
You can’t drive my car - a method to fingerprint individual driving styles in a sim-racing setting
The rapidly growing market of eSports has become a lucrative industry. Sim racing is one of the traditional sports that gain increasing popularity in eSports. With raising profits, lucrative sponsoring contracts, and competitive price money, incentives for fraud are also on the upraise. This is further facilitated by the COVID-19 pandemic, which led to the substitution of live eSport events by virtual formats. Particularly, in sim racing it becomes increasingly challenging to verify the contestants identity in virtual events. Here, we propose and evaluate a generic workflow to identify personal driving style by transforming raw racing data (including measurements extracted from the simulation software and connected simulator hardware) into a comparable representation (fingerprint). As data base we used an extensive collection of telemetry data recorded from the racing simulation Assetto Corsa on 3 immersive motion simulators. The set contains over 2,000 laps of seven player recorded in weekly sessions. The experiments demonstrate the feasibility to distinguish players using the proposed method on different tracks, car models, and motion simulators. Due to the extensive experiments, we could achieve a driver separation score of up to 89%. The raw data and the raw results are made publicly available on Github.
- 20 July 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 19, 1-9
You can’t drive my car - a method to fingerprint individual driving styles in a sim-racing setting
The rapidly growing market of eSports has become a lucrative industry. Sim racing is one of the traditional sports that gain increasing popularity in eSports. With raising profits, lucrative sponsoring contracts, and competitive price money, incentives for fraud are also on the upraise. This is further facilitated by the COVID-19 pandemic, which led to the substitution of live eSport events by virtual formats. Particularly, in sim racing it becomes increasingly challenging to verify the contestants identity in virtual events. Here, we propose and evaluate a generic workflow to identify personal driving style by transforming raw racing data (including measurements extracted from the simulation software and connected simulator hardware) into a comparable representation (fingerprint). As data base we used an extensive collection of telemetry data recorded from the racing simulation Assetto Corsa on 3 immersive motion simulators. The set contains over 2,000 laps of seven player recorded in weekly sessions. The experiments demonstrate the feasibility to distinguish players using the proposed method on different tracks, car models, and motion simulators. Due to the extensive experiments, we could achieve a driver separation score of up to 89%. The raw data and the raw results are made publicly available on Github.
- 20 July 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 20, 1-6
Data set for long term driving recording using immersive motion simulation in virtual racing - IMSRace
Research of driver behavior has become increasingly important for a multitude of stakeholders. However publicly available data sets are sparse. Most accessible corpora focus on driver behavior in real day to day traffic situations, with very little data being available for research in a competitive racing setting. This work presents a data set composed of recordings from different drivers using the virtual racing simulation software Assetto Corsa alongside three immersive motion simulator platforms. Altogether we recorded a data corpus consisting of more than 2200 laps from seven players, driven over eight weekly sessions. Our data set can be easily accessed and used with an additionally provided Python-library ims-lib (https://github.com/Mereep/1ib_ims) (Open Source /GPL3) and is accompanied with a repository (https://github.com/Mereep/ims-paper) specific for the data described here. The data set is freely accessible under Attribution-ShareAlike 4.0 International licence. Additional examples and material on how to access and use the data are included in the repositories.
- 20 July 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 20, 1-6
Data set for long term driving recording using immersive motion simulation in virtual racing - IMSRace
Research of driver behavior has become increasingly important for a multitude of stakeholders. However publicly available data sets are sparse. Most accessible corpora focus on driver behavior in real day to day traffic situations, with very little data being available for research in a competitive racing setting. This work presents a data set composed of recordings from different drivers using the virtual racing simulation software Assetto Corsa alongside three immersive motion simulator platforms. Altogether we recorded a data corpus consisting of more than 2200 laps from seven players, driven over eight weekly sessions. Our data set can be easily accessed and used with an additionally provided Python-library ims-lib (https://github.com/Mereep/1ib_ims) (Open Source /GPL3) and is accompanied with a repository (https://github.com/Mereep/ims-paper) specific for the data described here. The data set is freely accessible under Attribution-ShareAlike 4.0 International licence. Additional examples and material on how to access and use the data are included in the repositories.
- 6 July 2019
- Published by Springer Nature
- p. 295-302
Synthetic Ground Truth Generation for Object Recognition Evaluation
- 6 July 2019
- Published by Springer Nature
- p. 295-302
Synthetic Ground Truth Generation for Object Recognition Evaluation
- 6 July 2019
- Published by Springer Nature
- p. 190-197
Adaptation of Machine Learning Frameworks for Use in a Management Environment
- 6 July 2019
- Published by Springer Nature
- p. 190-197
Adaptation of Machine Learning Frameworks for Use in a Management Environment
- Open Access
- 1 July 2018
- Published by BCS Learning & Development Ltd
ALBIGen - A Framework to Create Managed and Distributed Digital Signage Systems Based on Laravel and iBeacons
Christian Roschke, Rico Thomanek, Robert Manthey, Maik Benndorf & Marc Ritter
- 1 July 2018
- Published by BCS Learning & Development Ltd
Open AccessALBIGen - A Framework to Create Managed and Distributed Digital Signage Systems Based on Laravel and iBeacons
Christian Roschke, Rico Thomanek, Robert Manthey, Maik Benndorf & Marc Ritter