Search

This paper presents CleanUMamba, a time-domain neural network architecture designed for real-time causal audio denoising directly applied to raw waveforms. CleanUMamba leverages a U-Net encoder-decoder structure, incorporating the Mamba state-space model in the bottleneck layer. By replacing conventional self-attention and LSTM mechanisms with Mamba, our architecture offers superior denoising performance while maintaining a constant memory footprint, enabling streaming operation. To enhance efficiency, we applied structured channel pruning, achieving an 8X reduction in model size without compromising audio quality. Our model demonstrates strong results in the Interspeech 2020 Deep Noise Suppression challenge. Specifically, CleanUMamba achieves a PESQ score of 2.42 and STOI of 95.1% with only 442K parameters and 468M MACs, matching or outperforming larger models in real-time performance. Code will be available at: https://github.com/lab-emi/CleanUMamba, Comment: 5 pages, 4 figures

Video annotation is a critical and time-consuming task in computer vision research and applications. This paper presents a novel annotation pipeline that uses pre-extracted features and dimensionality reduction to accelerate the temporal video annotation process. Our approach uses Hierarchical Stochastic Neighbor Embedding (HSNE) to create a multi-scale representation of video features, allowing annotators to efficiently explore and label large video datasets. We demonstrate significant improvements in annotation effort compared to traditional linear methods, achieving more than a 10x reduction in clicks required for annotating over 12 hours of video. Our experiments on multiple datasets show the effectiveness and robustness of our pipeline across various scenarios. Moreover, we investigate the optimal configuration of HSNE parameters for different datasets. Our work provides a promising direction for scaling up video annotation efforts in the era of video understanding.

In this paper, we jointly combine image classification and image denoising, aiming to enhance human perception of noisy images captured by edge devices, like low-light security cameras. In such settings, it is important to retain the ability of humans to verify the automatic classification decision and thus jointly denoise the image to enhance human perception. Since edge devices have little computational power, we explicitly optimize for efficiency by proposing a novel architecture that integrates the two tasks. Additionally, we alter a Neural Architecture Search (NAS) method, which searches for classifiers to search for the integrated model while optimizing for a target latency, classification accuracy, and denoising performance. The NAS architectures outperform our manually designed alternatives in both denoising and classification, offering a significant improvement to human perception. Our approach empowers users to construct architectures tailored to domains like medical imaging, surveillance systems, and industrial inspections., Comment: Accepted paper at the ECCV 2024 workshop on Advances in Image Manipulation (AIM)

A flexible job shop scheduling problem (FJSSP) poses a complex optimization task in modeling real-world process scheduling tasks with conflicting objectives. To tackle FJSSPs, approximation methods are employed to ensure solutions are within acceptable timeframes. Quantum Annealing, a metaheuristic leveraging quantum mechanical effects, demonstrates superior solution quality in a shorter time compared to classical algorithms. However, due to hardware limitations of quantum annealers, hybrid algorithms become essential for solving larger FJSSPs. This paper investigates the threshold problem sizes up to which quantum annealers are sufficient and when hybrid algorithms are required, highlighting the distribution of computing power in hybrid methods., Comment: 6 pages, 4 figures, 2 tables, accepted in 18th CIRP Conference on Intelligent Computation in Manufacturing Engineering

Flexible, self-organizing communication networks will be a key feature in the next mobile communication standard. Network-in-Network (NiN) is one important concept in 6G research, introducing sub-networks tailored to specific application requirements. These sub-networks may be dynamic, i.e., they may appear, disappear, or even move throughout the network. Moreover, sub-networks may operate within a shared frequency spectrum, thereby requiring coordination among them. We demonstrate the concept of Dynamic Spectrum Management (DSM) for future 6G networks that dynamically (re-)allocates spectrum according to active sub-networks in the shared spectrum domain. Resilient control plane connectivity between sub-networks and the DSM is provided by the self-organizing routing protocol KIRA, enabling the aforementioned coordination. This demonstration presents an integrated solution of the DSM concept, providing increased flexibility to support diverse industrial applications and their individual performance requirements simultaneously within the context of a cyber-physical production system (CPPS). For the sub-networks, we use specifically designed hardware for wireless real-time communication and couple them with a network emulation. By switching sub-networks on and off, one can see that the DSM dynamically manages the spectrum allocations for them and that KIRA provides the required connectivity., Comment: 5 pages, 3 figures

Protein characterization using nanopore-based devices promises to be a breakthrough method in basic research, diagnostics, and analytics. Current research includes the use of machine learning to achieve this task. In this work, a comprehensive statistical analysis of nanopore current signals is performed and demonstrated to be sufficient for classifying up to 42 peptides with over 70 % accuracy. Two sets of features, the statistical moments and the catch22 set, are compared both in their representations and after training small classifier neural networks. We demonstrate that complex features of the events, captured in both the catch22 set and the central moments, are key in classifying peptides with otherwise similar mean currents. These results highlight the efficacy of purely statistical analysis of nanopore data and suggest a path forward for more sophisticated classification techniques.

We present a functional-based approach to compute thermal expectation values for actions expressed in the $G-\Sigma$ formalism, applicable to any time sequence ordering. Utilizing this framework, we analyze the linear response to an electric field in various Sachdev-Ye-Kitaev (SYK) chains. We consider the SYK chain where each dot is a complex $q/2$-body interacting SYK model, and we allow for $r/2$-body nearest-neighbor hopping where $r=\kappa q$. We find exact analytical expressions in the large-$q$ limit for conductivities across all temperatures at leading order in $1/q$ for three cases, namely $\kappa = \{ 1/2, 1, 2\}$. When $\kappa = \{1/2, 1\}$, we observe linear-in-temperature $T$ resistivities at low temperatures, indicative of strange metal behavior. Conversely, when $\kappa = 2$, the resistivity diverges as a power law at low temperatures, namely as $1/T^2$, resembling insulating behavior. As $T$ increases, there is a crossover to Fermi liquid behavior ($\sim T^2$) at the minimum resistivity. Beyond this, another crossover occurs to strange metal behavior ($\sim T$). In comparison to previous linear-in-$T$ results in the literature, we also show that the resistivity behavior exists even below the MIR bound, indicating a true strange metal instead of a bad metal. In particular, we find for the $\kappa = 2$ case a smooth crossover from an insulating phase to a Fermi liquid behavior to a true strange metal and eventually becoming a bad metal as temperature increases. We extend and generalize previously known results on resistivities to all temperatures, do a comparative analysis across the three models where we highlight the universal features and invoke scaling arguments to create a physical picture out of our analyses. Remarkably, we find a universal maximum DC conductivity across all three models when the hopping coupling strength becomes large., Comment: 28 pages, 2 figures, published version

Recent studies have underscored the capabilities of natural imaging foundation models to serve as powerful feature extractors, even in a zero-shot setting for medical imaging data. Most commonly, a shallow multi-layer perceptron (MLP) is appended to the feature extractor to facilitate end-to-end learning and downstream prediction tasks such as classification, thus representing the de facto standard. However, as graph neural networks (GNNs) have become a practicable choice for various tasks in medical research in the recent past, we direct attention to the question of how effective GNNs are compared to MLP prediction heads for the task of 3D medical image classification, proposing them as a potential alternative. In our experiments, we devise a subject-level graph for each volumetric dataset instance. Therein latent representations of all slices in the volume, encoded through a DINOv2 pretrained vision transformer (ViT), constitute the nodes and their respective node features. We use public datasets to compare the classification heads numerically and evaluate various graph construction and graph convolution methods in our experiments. Our findings show enhancements of the GNN in classification performance and substantial improvements in runtime compared to an MLP prediction head. Additional robustness evaluations further validate the promising performance of the GNN, promoting them as a suitable alternative to traditional MLP classification heads. Our code is publicly available at: https://github.com/compai-lab/2024-miccai-grail-kiechle, Comment: Accepted at MICCAI 2024 - GRAIL Workshop

In this work, we introduce Progressive Growing of Patch Size, a resource-efficient implicit curriculum learning approach for dense prediction tasks. Our curriculum approach is defined by growing the patch size during model training, which gradually increases the task's difficulty. We integrated our curriculum into the nnU-Net framework and evaluated the methodology on all 10 tasks of the Medical Segmentation Decathlon. With our approach, we are able to substantially reduce runtime, computational costs, and CO2 emissions of network training compared to classical constant patch size training. In our experiments, the curriculum approach resulted in improved convergence. We are able to outperform standard nnU-Net training, which is trained with constant patch size, in terms of Dice Score on 7 out of 10 MSD tasks while only spending roughly 50% of the original training runtime. To the best of our knowledge, our Progressive Growing of Patch Size is the first successful employment of a sample-length curriculum in the form of patch size in the field of computer vision. Our code is publicly available at https://github.com/compai-lab/2024-miccai-fischer., Comment: Accepted at MICCAI2024; Changes for Camera-Ready-Version for MICCAI2024 (missing in this arxiv submission): Replaced T-Test with Wilcoxon Signed Ranked Test, as DSC samples are not normally distributed => now only significant improvements and no significant decreases in performance for PGPS/PGPS+

Pathological lymph node delineation is crucial in cancer diagnosis, progression assessment, and treatment planning. The MICCAI 2023 Lymph Node Quantification Challenge published the first public dataset for pathological lymph node segmentation in the mediastinum. As lymph node annotations are expensive, the challenge was formed as a weakly supervised learning task, where only a subset of all lymph nodes in the training set have been annotated. For the challenge submission, multiple methods for training on these weakly supervised data were explored, including noisy label training, loss masking of unlabeled data, and an approach that integrated the TotalSegmentator toolbox as a form of pseudo labeling in order to reduce the number of unknown voxels. Furthermore, multiple public TCIA datasets were incorporated into the training to improve the performance of the deep learning model. Our submitted model achieved a Dice score of 0.628 and an average symmetric surface distance of 5.8~mm on the challenge test set. With our submitted model, we accomplished third rank in the MICCAI2023 LNQ challenge. A finding of our analysis was that the integration of all visible, including non-pathological, lymph nodes improved the overall segmentation performance on pathological lymph nodes of the test set. Furthermore, segmentation models trained only on clinically enlarged lymph nodes, as given in the challenge scenario, could not generalize to smaller pathological lymph nodes. The code and model for the challenge submission are available at \url{https://gitlab.lrz.de/compai/MediastinalLymphNodeSegmentation}., Comment: Accepted for publication at the Journal of Machine Learning for Biomedical Imaging (MELBA) https://melba-journal.org/2024:008

Background: The aim of this study was to investigate the role of clinical, dosimetric and pretherapeutic magnetic resonance imaging (MRI) features for lesion-specific outcome prediction of stereotactic radiotherapy (SRT) in patients with brain metastases from malignant melanoma (MBM). Methods: In this multicenter, retrospective analysis, we reviewed 517 MBM from 130 patients treated with SRT (single fraction or hypofractionated). For each gross tumor volume (GTV) 1576 radiomic features (RF) were calculated (788 each for the GTV and for a 3 mm margin around the GTV). Clinical parameters, radiation dose and RF from pretherapeutic contrast-enhanced T1-weighted MRI from different institutions were evaluated with a feature processing and elimination pipeline in a nested cross-validation scheme. Results: Seventy-two (72) of 517 lesions (13.9%) showed a local failure (LF) after SRT. The processing pipeline showed clinical, dosimetric and radiomic features providing information for LF prediction. The most prominent ones were the correlation of the gray level co-occurrence matrix of the margin (hazard ratio (HR): 0.37, confidence interval (CI): 0.23-0.58) and systemic therapy before SRT (HR: 0.55, CI: 0.42-0.70). The majority of RF associated with LF was calculated in the margin around the GTV. Conclusions: Pretherapeutic MRI based RF connected with lesion-specific outcome after SRT could be identified, despite multicentric data and minor differences in imaging protocols. Image data analysis of the surrounding metastatic environment may provide therapy-relevant information with the potential to further individualize radiotherapy strategies.

We propose a method to construct representative price profiles of the day-ahead (DA) and the intraday (ID) electricity spot markets and use this method to provide examples of ready-to-use price data sets. In contrast to common scenario generation approaches, the method is deterministic and relies on a small number of degrees of freedom, with the aim to be well defined and easy to use. We thereby target an enhanced comparability of future research studies on demand-side management and energy cost optimization. We construct the price profiles based on historical time series from the spot markets of interest, e.g., European Power Exchange (EPEX) spot. To this end, we extract key price components from the data while also accounting for known dominant mechanisms in the price variation. Further, the method is able to preserve key statistical features of the historical data (e.g., mean and standard deviation) when constructing the benchmark profile. Finally, our approach ensures comparability of ID and DA price profiles by design, as their cumulative (integral) price can be made identical if needed., Comment: Supplementary information (SI) included; Manuscript: 27 pages, 9 figures, 4 tables; SI: 7 pages, 5 figures, 2 tables

CNNs and computational models of biological vision share some fundamental principles, which opened new avenues of research. However, fruitful cross-field research is hampered by conventional CNN architectures being based on spatially and depthwise discrete representations, which cannot accommodate certain aspects of biological complexity such as continuously varying receptive field sizes and dynamics of neuronal responses. Here we propose deep continuous networks (DCNs), which combine spatially continuous filters, with the continuous depth framework of neural ODEs. This allows us to learn the spatial support of the filters during training, as well as model the continuous evolution of feature maps, linking DCNs closely to biological models. We show that DCNs are versatile and highly applicable to standard image classification and reconstruction problems, where they improve parameter and data efficiency, and allow for meta-parametrization. We illustrate the biological plausibility of the scale distributions learned by DCNs and explore their performance in a neuroscientifically inspired pattern completion task. Finally, we investigate an efficient implementation of DCNs by changing input contrast., Comment: Presented at ICML 2021

We use Koopman theory for data-driven model reduction of nonlinear dynamical systems with controls. We propose generic model structures combining delay-coordinate encoding of measurements and full-state decoding to integrate reduced Koopman modeling and state estimation. We present a deep-learning approach to train the proposed models. A case study demonstrates that our approach provides accurate control models and enables real-time capable nonlinear model predictive control of a high-purity cryogenic distillation column.

It has been known that the large-$q$ complex SYK model falls under the same universality class as that of van der Waals (mean-field) and saturates the Maldacena-Shenker-Stanford bound, both features shared by various black holes. This makes the SYK model a useful tool in probing the fundamental nature of quantum chaos and holographic duality. This work establishes the robustness of this shared universality class and chaotic properties for SYK-like models by extending to a system of coupled large-$q$ complex SYK models of different orders. We provide a detailed derivation of thermodynamic properties, specifically the critical exponents for an observed phase transition, as well as dynamical properties, in particular the Lyapunov exponent, via the out-of-time correlator calculations. Our analysis reveals that, despite the introduction of an additional scaling parameter through interaction strength ratios, the system undergoes a continuous phase transition at low temperatures, similar to that of the single SYK model. The critical exponents align with the Landau-Ginzburg (mean-field) universality class, shared with van der Waals gases and various AdS black holes. Furthermore, we demonstrate that the coupled SYK system remains maximally chaotic in the large-$q$ limit at low temperatures, adhering to the Maldacena-Shenker-Stanford bound, a feature consistent with the single SYK model. These findings establish robustness and open avenues for broader inquiries into the universality and chaos in complex quantum systems. We provide a detailed outlook for future work by considering the "very" low-temperature regime, where we discuss relations with the Hawking-Page phase transition observed in the holographic dual black holes. We present preliminary calculations and discuss the possible follow-ups that might be taken to make the connection robust., Comment: 20 pages, 10 figures, 1 table, same as published version but in a different formatting style

Flexible Job Shop Scheduling (FJSSP) is a complex optimization problem crucial for real-world process scheduling in manufacturing. Efficiently solving such problems is vital for maintaining competitiveness. This paper introduces Quantum Annealing-based solving algorithm (QASA) to address FJSSP, utilizing quantum annealing and classical techniques. QASA optimizes multi-criterial FJSSP considering makespan, total workload, and job priority concurrently. It employs Hamiltonian formulation with Lagrange parameters to integrate constraints and objectives, allowing objective prioritization through weight assignment. To manage computational complexity, large instances are decomposed into subproblems, and a decision logic based on bottleneck factors is used. Experiments on benchmark problems show QASA, combining tabu search, simulated annealing, and Quantum Annealing, outperforms a classical solving algorithm (CSA) in solution quality (set coverage and hypervolume ratio metrics). Computational efficiency analysis indicates QASA achieves superior Pareto solutions with a reasonable increase in computation time compared to CSA., Comment: 21 pages, 6 figures, 2 tables