6,937 Hits in 1.2 sec

Prune and Replace NAS [article]

Kevin Alexander Laube, Andreas Zell
2019 arXiv   pre-print
While recent NAS algorithms are thousands of times faster than the pioneering works, it is often overlooked that they use fewer candidate operations, resulting in a significantly smaller search space. We present PR-DARTS, a NAS algorithm that discovers strong network configurations in a much larger search space and a single day. A small candidate operation pool is used, from which candidates are progressively pruned and replaced with better performing ones. Experiments on CIFAR-10 and CIFAR-100
more » ... achieve 2.51% and 15.53% test error, respectively, despite searching in a space where each cell has 150 times as many possible configurations than in the DARTS baseline. Code is available at
arXiv:1906.07528v2 fatcat:4tqj5crerbgube6thlvtrmg7hi

Context-based generation of kinetic equations with SBMLsqueezer 1.3

Andreas Dräger, Andreas Dräger, Sandra Nitschmann, Alexander Dörr, Johannes Eichner, Michael Ziller, Andreas Zell
2010 Nature Precedings  
Generalized mass-action equation Reversible Michaelis-Menten equation with inhibition Convenience rate law Langevin equation 2 | Andreas Dräger et al.  ...  : Posted 9 Oct 2010 | Andreas Dräger et al. © 2010 Universität Tübingen Nature Precedings : doi:10.1038/npre.2010.4983.1 : Posted 9 Oct 2010 | Andreas Dräger et al. © 2010 Universität Tübingen  ...  npre.2010.4983.1 : Posted 9 Oct 2010 | Andreas Dräger et al. © 2010 Universität Tübingen • time (in seconds) Nature Precedings : doi:10.1038/npre.2010.4983.1 : Posted 9 Oct 2010 | Andreas Dräger  ... 
doi:10.1038/npre.2010.4983 fatcat:rev6pwz3xbgilffxl4rby23ehm

Context-based generation of kinetic equations with SBMLsqueezer 1.3

Andreas Dräger, Andreas Dräger, Sandra Nitschmann, Alexander Dörr, Johannes Eichner, Michael Ziller, Andreas Zell
2010 Nature Precedings  
Generalized mass-action equation Reversible Michaelis-Menten equation with inhibition Convenience rate law Langevin equation 2 | Andreas Dräger et al.  ...  : Posted 9 Oct 2010 | Andreas Dräger et al. © 2010 Universität Tübingen Nature Precedings : doi:10.1038/npre.2010.4983.1 : Posted 9 Oct 2010 | Andreas Dräger et al. © 2010 Universität Tübingen  ...  npre.2010.4983.1 : Posted 9 Oct 2010 | Andreas Dräger et al. © 2010 Universität Tübingen • time (in seconds) Nature Precedings : doi:10.1038/npre.2010.4983.1 : Posted 9 Oct 2010 | Andreas Dräger  ... 
doi:10.1038/npre.2010.4983.1 fatcat:d56mhjcjgzegzhnmthrp4v7sau

Conditional super-network weights [article]

Kevin Alexander Laube, Andreas Zell
2021 arXiv   pre-print
Modern Neural Architecture Search methods have repeatedly broken state-of-the-art results for several disciplines. The super-network, a central component of many such methods, enables quick estimates of accuracy or loss statistics for any architecture in the search space. They incorporate the network weights of all candidate architectures and can thus approximate specific ones by applying the respective operations. However, this design ignores potential dependencies between consecutive
more » ... s. We extend super-networks with conditional weights that depend on combinations of choices and analyze their effect. Experiments in NAS-Bench 201 and NAS-Bench-Macro-based search spaces show improvements in the architecture selection and that the resource overhead is nearly negligible for sequential network designs.
arXiv:2104.11522v2 fatcat:b3w5jr74lfd5znrc6ozdogrrcy

BowTieBuilder: modeling signal transduction pathways

Jochen Supper, Lucía Spangenberg, Hannes Planatscher, Andreas Dräger, Adrian Schröder, Andreas Zell
2009 BMC Systems Biology  
Sensory proteins react to changing environmental conditions by transducing signals into the cell. These signals are integrated into core proteins that activate downstream target proteins such as transcription factors (TFs). This structure is referred to as a bow tie, and allows cells to respond appropriately to complex environmental conditions. Understanding this cellular processing of information, from sensory proteins (e.g., cell-surface proteins) to target proteins (e.g., TFs) is important,
more » ... et for many processes the signaling pathways remain unknown. Results: Here, we present BowTieBuilder for inferring signal transduction pathways from multiple source and target proteins. Given protein-protein interaction (PPI) data signaling pathways are assembled without knowledge of the intermediate signaling proteins while maximizing the overall probability of the pathway. To assess the inference quality, BowTieBuilder and three alternative heuristics are applied to several pathways, and the resulting pathways are compared to reference pathways taken from KEGG. In addition, BowTieBuilder is used to infer a signaling pathway of the innate immune response in humans and a signaling pathway that potentially regulates an underlying gene regulatory network. Conclusion: We show that BowTieBuilder, given multiple source and/or target proteins, infers pathways with satisfactory recall and precision rates and detects the core proteins of each pathway.
doi:10.1186/1752-0509-3-67 pmid:19566957 pmcid:PMC2712453 fatcat:c3cc7ysmxbgpnfdep6qdhh6dpe

Gaze-based Object Detection in the Wild [article]

Daniel Weber, Wolfgang Fuhl, Andreas Zell, Enkelejda Kasneci
2022 arXiv   pre-print
In human-robot collaboration, one challenging task is to teach a robot new yet unknown objects. Thereby, gaze can contain valuable information. We investigate if it is possible to detect objects (object or no object) from gaze data and determine their bounding box parameters. For this purpose, we explore different sizes of temporal windows, which serve as a basis for the computation of heatmaps, i.e., the spatial distribution of the gaze data. Additionally, we analyze different grid sizes of
more » ... se heatmaps, and various machine learning techniques are applied. To generate the data, we conducted a small study with five subjects who could move freely and thus, turn towards arbitrary objects. This way, we chose a scenario for our data collection that is as realistic as possible. Since the subjects move while facing objects, the heatmaps also contain gaze data trajectories, complicating the detection and parameter regression.
arXiv:2203.15651v1 fatcat:l4umb6q725g2bi5eejh5syqwu4

Collaborative Localization and Tracking with Minimal Infrastructure [article]

Yanjun Cao, David St-Onge, Andreas Zell, Giovanni Beltrame
2019 arXiv   pre-print
Zell is with the Wilhelm-Schickard-Institute for Computer Science, University of Tübingen, Germany, e-mail: ( to share the users experience on social networks.  ... 
arXiv:1905.03247v1 fatcat:fjcy7cjuubeubivpt3tpzywuea

Optimal assignment methods for ligand-based virtual screening

Andreas Jahn, Georg Hinselmann, Nikolas Fechner, Andreas Zell
2009 Journal of Cheminformatics  
Ligand-based virtual screening experiments are an important task in the early drug discovery stage. An ambitious aim in each experiment is to disclose active structures based on new scaffolds. To perform these "scaffold-hoppings" for individual problems and targets, a plethora of different similarity methods based on diverse techniques were published in the last years. The optimal assignment approach on molecular graphs, a successful method in the field of quantitative structure-activity
more » ... nships, has not been tested as a ligand-based virtual screening method so far. Results: We evaluated two already published and two new optimal assignment methods on various data sets. To emphasize the "scaffold-hopping" ability, we used the information of chemotype clustering analyses in our evaluation metrics. Comparisons with literature results show an improved early recognition performance and comparable results over the complete data set. A new method based on two different assignment steps shows an increased "scaffold-hopping" behavior together with a good early recognition performance. Conclusion: The presented methods show a good combination of chemotype discovery and enrichment of active structures. Additionally, the optimal assignment on molecular graphs has the advantage to investigate and interpret the mappings, allowing precise modifications of internal parameters of the similarity measure for specific targets. All methods have low computation times which make them applicable to screen large data sets.
doi:10.1186/1758-2946-1-14 pmid:20150995 pmcid:PMC2820492 fatcat:7ho5qwxl6jh57en6o45inyhbnm

Seeing Implicit Neural Representations as Fourier Series [article]

Nuri Benbarka, Timon Höfer, Hamd ul-moqeet Riaz, Andreas Zell
2021 arXiv   pre-print
Implicit Neural Representations (INR) use multilayer perceptrons to represent high-frequency functions in low-dimensional problem domains. Recently these representations achieved state-of-the-art results on tasks related to complex 3D objects and scenes. A core problem is the representation of highly detailed signals, which is tackled using networks with periodic activation functions (SIRENs) or applying Fourier mappings to the input. This work analyzes the connection between the two methods
more » ... shows that a Fourier mapped perceptron is structurally like one hidden layer SIREN. Furthermore, we identify the relationship between the previously proposed Fourier mapping and the general d-dimensional Fourier series, leading to an integer lattice mapping. Moreover, we modify a progressive training strategy to work on arbitrary Fourier mappings and show that it improves the generalization of the interpolation task. Lastly, we compare the different mappings on the image regression and novel view synthesis tasks. We confirm the previous finding that the main contributor to the mapping performance is the size of the embedding and standard deviation of its elements.
arXiv:2109.00249v1 fatcat:p5stx7hsn5avxoeejxamwvaply

MobileStereoNet: Towards Lightweight Deep Networks for Stereo Matching [article]

Faranak Shamsafar, Samuel Woerz, Rafia Rahim, Andreas Zell
2021 arXiv   pre-print
Recent methods in stereo matching have continuously improved the accuracy using deep models. This gain, however, is attained with a high increase in computation cost, such that the network may not fit even on a moderate GPU. This issue raises problems when the model needs to be deployed on resource-limited devices. For this, we propose two light models for stereo vision with reduced complexity and without sacrificing accuracy. Depending on the dimension of cost volume, we design a 2D and a 3D
more » ... del with encoder-decoders built from 2D and 3D convolutions, respectively. To this end, we leverage 2D MobileNet blocks and extend them to 3D for stereo vision application. Besides, a new cost volume is proposed to boost the accuracy of the 2D model, making it performing close to 3D networks. Experiments show that the proposed 2D/3D networks effectively reduce the computational expense (27%/95% and 72%/38% fewer parameters/operations in 2D and 3D models, respectively) while upholding the accuracy. Our code is available at
arXiv:2108.09770v1 fatcat:6adqae5l5bgafhottjnmedh7gm

Score refinement for confidence-based 3D multi-object tracking [article]

Nuri Benbarka, Jona Schröder, Andreas Zell
2021 arXiv   pre-print
. * UNPUBLISHED WORK Nuri Benbarka, Jona Schröder, and Andreas Zell are with the cognitive systems group, University of Tübingen, Sand 1, Tübingen, Germany  ... 
arXiv:2107.04327v1 fatcat:hohqo2ojcvfflneb4mmjg477om

The EvA2 Optimization Framework [chapter]

Marcel Kronfeld, Hannes Planatscher, Andreas Zell
2010 Lecture Notes in Computer Science  
We present EvA2, a comprehensive metaheuristic optimiza-
doi:10.1007/978-3-642-13800-3_27 fatcat:26htizwwlngq5ornuto74lcmqm

ShuffleNASNets: Efficient CNN models through modified Efficient Neural Architecture Search [article]

Kevin Alexander Laube, Andreas Zell
2018 arXiv   pre-print
Neural network architectures found by sophistic search algorithms achieve strikingly good test performance, surpassing most human-crafted network models by significant margins. Although computationally efficient, their design is often very complex, impairing execution speed. Additionally, finding models outside of the search space is not possible by design. While our space is still limited, we implement undiscoverable expert knowledge into the economic search algorithm Efficient Neural
more » ... ure Search (ENAS), guided by the design principles and architecture of ShuffleNet V2. While maintaining baseline-like 2.85% test error on CIFAR-10, our ShuffleNASNets are significantly less complex, require fewer parameters, and are two times faster than the ENAS baseline in a classification task. These models also scale well to a low parameter space, achieving less than 5% test error with little regularization and only 236K parameters.
arXiv:1812.02975v1 fatcat:5r7aoq3iq5h57kw76zgnxouqfm

Comprehensive Analysis of the Object Detection Pipeline on UAVs [article]

Leon Amadeus Varga, Sebastian Koch, Andreas Zell
2022 arXiv   pre-print
An object detection pipeline comprises a camera that captures the scene and an object detector that processes these images. The quality of the images directly affects the performance of the object detector. Many works nowadays focus either on improving the image quality or improving the object detection models independently, but neglect the importance of joint optimization of the two subsystems. In this paper, we first empirically analyze the influence of seven parameters (quantization,
more » ... ion, resolution, color model, image distortion, gamma correction, additional channels) in remote sensing applications. For our experiments, we utilize three UAV data sets from different domains and a mixture of large and small state-of-the-art object detector models to provide an extensive evaluation of the influence of the pipeline parameters. Additionally, we realize an object detection pipeline prototype on an embedded platform for an UAV and give a best practice recommendation for building object detection pipelines based on our findings. We show that not all parameters have an equal impact on detection accuracy and data throughput, and that by using a suitable compromise between parameters we are able to improve detection accuracy for lightweight object detection models, while keeping the same data throughput.
arXiv:2203.00306v1 fatcat:c35femic35dcfdqa5xfy7woh6a

ProDGe: investigating protein-protein interactions at the domain level

Finja Büchel, Finja Büchel, Clemens Wrzodek, Florian Mittag, Andreas Dräger, Adrian Schröder, Andreas Zell
2011 Nature Precedings  
An important goal of systems biology is the identification and investigation of known and predicted proteinprotein interactions to obtain more information about new cellular pathways and processes. Proteins interact via domains, thus it is important to know which domains a protein contains and which domains interact with each other. Here we present the Java TM program ProDGe (Protein Domain Gene), which visualizes existing and suggests novel domain-domain interactions and protein-protein
more » ... tions at the domain level. The comprehensive dataset behind ProDGe consists of protein, domain and interaction information for both layers, collected and combined appropriately from UniProt, Pfam, DOMINE and IntAct. Based on known domain interactions, ProDGe suggests novel protein interactions and assigns them to four confidence classes, depending on the reliability of the underlying domain interaction. Furthermore, ProDGe is able to identify potential homologous interaction partners in other species, which is particularly helpful when investigating poorly annotated species. We further evaluated and compared experimentally identified protein interactions from IntAct with domain interactions from DOMINE for six species and noticed that 31.13% of all IntAct protein interactions in all six species can be mapped to the actual interacting domains. ProDGe and a comprehensive documentation are freely available at
doi:10.1038/npre.2011.6188 fatcat:x6vcfjx2e5gcxfsx5ebixwffui
« Previous Showing results 1 — 15 out of 6,937 results