Intelligent Autonomous Systems (INTAS)

She will work in the project "AI-DPA: Analyse und Interpretation von unstrukturierten Daten und Prozessen in zwei- und dreidimensionalen Anwendungsszenarien mit Machine Learning"

Although I already started in Koblenz in 2022, I will now give my inaugural lecture with the title "Künstliche Intelligenz hat viele Dimensionen: Die Magie hochdimensionaler Vektorrepräsentationen für Roboter und andere autonome Systeme" (the lecture will be in German).

Anyone interested is warmly invited!

February 1, 18:00-20:00, Room M001, University Campus

A joint team from our group and researchers from Sweden and Australia has won the "My Seizure Gauge'' Challenge on forecasting epileptic seizures from non-cerebral signals. A short summary of the competition and our winning approach is provided in a paper that has been accepted for publication in Nature Machine Intelligence in the Challenge Accepted track. Congratulations to Kenny Schlegel who has led this team! The paper is a joint work of our team and the challenge organizers.

Kenny Schlegel, Denis Kleyko, Benjamin H. Brinkmann, Ewan S. Nurse, Ross W. Gayler, Peer Neubert (2024). Lessons from the “My Seizure Gauge” Challenge on Forecasting Epileptic Seizures from Non-Cerebral Signals. Nature Machine Intelligence Challenge Accepted (to appear)

The paper "Local positional graphs and attentive local features for a data and runtime-efficient hierarchical place recognition pipeline" by Fangming Yuan, Stefan Schubert, Peter Protzel, and Peer Neubert has been accepted for publication in the IEEE RA-L journal.

Abstract: Large-scale applications of Visual Place Recognition (VPR) require computationally efficient approaches. Further, a well-balanced combination of data-based and training-free approaches can decrease the required amount of training data and effort and can reduce the influence of distribution shifts between the training and application phases. This paper proposes a runtime and data-efficient hierarchical VPR pipeline that extends existing approaches and presents novel ideas. There are three main contributions: First, we propose Local Positional Graphs (LPG), a training-free and runtime-efficient approach to encode spatial context information of local image features. LPG can be combined with existing local feature detectors and descriptors and considerably improves the image-matching quality compared to existing techniques in our experiments. Second, we present Attentive Local SPED (ATLAS), an extension of our previous local features approach with an attention module that improves the feature quality while maintaining high data efficiency. The influence of the proposed modifications is evaluated in an extensive ablation study. Third, we present a hierarchical pipeline that exploits hyperdimensional computing to use the same local features as holistic HDC-descriptors for fast candidate selection and for candidate reranking. We combine all contributions in a runtime and data-efficient VPR pipeline that shows benefits over the state-of-the-art method Patch-NetVLAD on a large collection of standard place recognition datasets with 15x better performance in VPR accuracy, 54x faster feature comparison speed, and 27x less descriptor storage occupancy, making our method promising for real-world high-performance large-scale VPR in changing environments. Code will be made available with publication of this paper.

Our colleague Mark O. Mints gave a presentation on the use of robot simulation in teaching at the first ROScon in the DACH region. The project "VacuSim" is available on GitLab.

Further information can be found on the official ROScon DE website. Video recordings will be available there shortly.

Stefan Schubert - an external PhD student - successfully defended his PhD thesis at Chemnitz University of Technology with highest honours. Congratulations!

The paper "FETCH: A Memory-Efficient Replay Approach for Continual Learning in Image Classification" was accepted at IDEAL conference.

Abstract: Class-incremental continual learning is an important area of research, as static deep learning methods fail to adapt to changing tasks and data distributions. In previous works, promising results were achieved using replay and compressed replay techniques. In the field of regular replay, GDumb [23] achieved outstanding results but requires a large amount of memory. This problem can be addressed by compressed replay techniques. The goal of this work is to evaluate compressed replay in the pipeline of GDumb. We propose FETCH, a two-stage compression approach. First, the samples from the continual datastream are encoded by the early layers of a pre-trained neural network. Second, the samples are compressed before being stored in the episodic memory. Following GDumb, the remaining classification head is trained from scratch using only the decompressed samples from the reply memory. We evaluate FETCH in different scenarios and show that this approach can increase accuracy on CIFAR10 and CIFAR100. In our experiments, simple compression methods (e.g., quantization of tensors) outperform deep autoencoders. In the future, FETCH could serve as a baseline for benchmarking compressed replay learning in constrained memory scenarios.

The KI-Forschungskolleg "AI-DPA: Analyse und Interpretation von unstrukturierten Daten und Prozessen in zwei- und dreidimensionalen Anwendungsszenarien mit Machine Learning" in cooperation with University of Applied Science Mainz started!