Jonas Krämer - Internet Archive Scholar

Normalizing flows are a promising tool for modeling probability distributions in physical systems. While state-of-the-art flows accurately approximate distributions and energies, applications in physics additionally require smooth energies to compute forces and higher-order derivatives. Furthermore, such densities are often defined on non-trivial topologies. A recent example are Boltzmann Generators for generating 3D-structures of peptides and small proteins. These generative models leverage

more »

... space of internal coordinates (dihedrals, angles, and bonds), which is a product of hypertori and compact intervals. In this work, we introduce a class of smooth mixture transformations working on both compact intervals and hypertori. Mixture transformations employ root-finding methods to invert them in practice, which has so far prevented bi-directional flow training. To this end, we show that parameter gradients and forces of such inverses can be computed from forward evaluations via the inverse function theorem. We demonstrate two advantages of such smooth flows: they allow training by force matching to simulation data and can be used as potentials in molecular dynamics simulations.

arXiv:2110.00351v2 fatcat:gl2blcawjjfhlmcntkfsmh5owm

Open Access Multiple Versions

Automatic tool detection from surgical imagery has a multitude of useful applications, such as real-time computer assistance for the surgeon. Using the successful residual network architecture, a system that can distinguish 21 different tools in cataract surgery videos is created. The videos are provided as part of the 2017 CATARACTS challenge and pose difficulties found in many real-world datasets, for example a strong class imbalance. The construction of the detection system is guided by a

more »

... e array of experiments that explore different design decisions.

arXiv:1805.05760v1 fatcat:dskky3kitffprf66kyw3fy2zjm

Kramer [10] uses a (1 + 1) − EA to optimize the hyperparameters defining a convolutional highway network. Suganuma et al. ... Therefore, a niching approach adapted from Kramer [10] is implemented. There is a random chance η to follow solutions that are initially worse. ...

arXiv:1806.08099v2 fatcat:67nraqpvabebznhxr75hsadctq

Multiple Versions

In this study, we investigated the relationship between changes in demand-side determinants and changes in hospital admissions. We used longitudinal market-wide data, including a novel detailed measure of population morbidity. To assess the effect of ageing, we interacted age with shifts in the population structure for both the surviving population and the population in their last year of life. We used fixed effects models and addressed the endogeneity of morbidity with instrumental variables.

more »

... e found that changes in morbidity had the largest impact on changes in hospital admissions. Changes in the size of the surviving population had the second largest impact, which differed substantially across the age spectrum. There was a large response in admissions to changes in the size of the population aged 60-79 years. The end-of-life effect had the smallest impact and began to play a greater role only in the population aged 80 years and older. In many studies, end of life presumably approximates high morbidity. Our results demonstrated robustness in several tests. We performed estimations in separate major diagnostic categories and included changes in personal preferences. We argue that the determinants included in our estimations capture the vast majority of change on the demand side. Taken together, our findings provide evidence that these determinants explain one-fifth of changes in hospital admissions.

doi:10.1007/s10198-019-01033-6 pmid:30739296 pmcid:PMC6602979 fatcat:s2glqvyqeja6zdzoonmdjn6ok4

Kramer [10] uses a (1 + 1) − EA to optimize the hyperparameters defining a convolutional highway network. Suganuma et al. ... Therefore, a niching approach adapted from Kramer [10] is implemented. There is a random chance η to follow solutions that are initially worse. ...

doi:10.1007/978-3-319-99259-4_34 fatcat:b5ieeqyho5f53l4g4z3uayhjiy

The real part of the refractive index (RI) of aqueous solutions of human hemoglobin is computed from their absorption spectra in the wavelength range 250 nm - 1100 nm using the Kramers-Kronig (KK) relations ... Real Part of Refractive Index by Kramers-Kronig Relations B.1. ... Kramers-Kronig Relations The complex refractive index is a linear, causal responsefunction to an incident wave. Hence its real and imaginary part are related by KK relations. ...

doi:10.1364/ao.55.008951 pmid:27828301 fatcat:wkntp56fszbqxdew6yeqdwrfzq

Multiple Versions

Stochastic gradient descent is the most prevalent algorithm to train neural networks. However, other approaches such as evolutionary algorithms are also applicable to this task. Evolutionary algorithms bring unique trade-offs that are worth exploring, but computational demands have so far restricted exploration to small networks with few parameters. We implement an evolutionary algorithm that executes entirely on the GPU, which allows to efficiently batch-evaluate a whole population of

more »

... Within this framework, we explore the limited evaluation evolutionary algorithm for neural network training and find that its batch evaluation idea comes with a large accuracy trade-off. In further experiments, we explore crossover operators and find that unprincipled random uniform crossover performs extremely well. Finally, we train a network with 92k parameters on MNIST using an EA and achieve 97.6 % test accuracy compared to 98 % test accuracy on the same network trained with Adam. Code is available at https

doi:10.1007/978-3-030-00111-7_23 fatcat:jdgvhz5qhbghha5c63l7bpujdu

We show that some care should be exercised when inferring true unconditional correlations from observed conditional correlations, which is a frequent problem in empirical finance and elsewhere. We give a general formula for the relationship between the two and demonstrate its importance in the context of the bivariate tdistribution.

doi:10.1016/j.econlet.2010.06.003 fatcat:dwzdw3garnh27oeqv4yfbcsui4

Symmetry is rarely found on cellular surfaces. An exception is the brush border of microvilli, which are essential for the proper function of transport epithelia. In a healthy intestine, they appear densely packed as a 2D-hexagonal lattice. For in vitro testing of intestinal transport the cell line Caco-2 has been established. As reported by electron microscopy, their microvilli arrange primarily in clusters developing secondly into a 2D-hexagonal lattice. Here, atomic force microscopy (AFM)

more »

... employed under aqueous buffer conditions on Caco-2 cells, which were cultivated on permeable filter membranes for optimum differentiation. For analysis, the exact position of each microvillus was detected by computer vision; subsequent Fourier transformation yielded the type of 2D-lattice. It was confirmed, that Caco-2 cells can build a hexagonal lattice of microvilli and form clusters. Moreover, a second type of arrangement was discovered, namely a rhombic lattice, which appeared at sub-maximal densities of microvilli with (29 ± 4) microvilli / μm 2 . Altogether, the findings indicate the existence of a yet undescribed pattern in cellular organization. Citation: Franz J, Grünebaum J, Schäfer M, Mulac D, Rehfeldt F, Langer K, et al. (2018) Rhombic organization of microvilli domains found in a cell model of the human intestine. PLoS ONE 13(1): e0189970. https://doi.org/10.

doi:10.1371/journal.pone.0189970 pmid:29320535 pmcid:PMC5761853 fatcat:qz6e5wiwqbfdfnfglvyo4aptnm

DOAJ

In deep multi-task learning, weights of task-specific networks are shared between tasks to improve performance on each single one. Since the question, which weights to share between layers, is difficult to answer, human-designed architectures often share everything but a last task-specific layer. In many cases, this simplistic approach severely limits performance. Instead, we propose an algorithm to learn the assignment between a shared set of weights and task-specific layers. To optimize the

more »

... n-differentiable assignment and at the same time train the differentiable weights, learning takes place via a combination of natural evolution strategy and stochastic gradient descent. The end result are task-specific networks that share weights but allow independent inference. They achieve lower test errors than baselines and methods from literature on three multi-task learning datasets.

arXiv:2003.10159v1 fatcat:m72kbv3b6nayvfyhj4ri3hitlm

Angiogenesis is essential for tumor growth and progression. It has been demonstrated that tumor growth beyond a size 1 to 2 mm 3 requires the induction of new vessels. Angiogenesis is regulated by several endogenous stimulators and inhibitors of endothelial cell migration, proliferation and tube formation. Under physiological conditions these mediators of endothelial cell growth are in balance and vessel growth is limited. In fact, within the angiogenic balance endothelial cell turnover is

more »

... cient to maintain a functional vascular wall but does not allow vessel growth. Tumor growth an progression has successfully been correlated to the serum concentration of angiogenic mediators. Furthermore, the vascular density of tumor tissues could be correlated to the clinical course of the disease in several tumor entities. Within the last years several new mediators of endothelial cell growth have been isolated e.g. angiopoietin 1, angiopoietin 2, midkine, pleiotropin, leptin and maspin. In this review we discuss the mechanisms leading to tumor angiogenesis and describe some of the newer mediators of endothelial cell stimulation and inhibition.

doi:10.1111/j.1582-4934.2000.tb00125.x pmid:12067460 fatcat:t5zsdzyvyffvtkf53bdllfacoa

DOAJ

Zusammenfassung Die Beurteilung der Erektionsfähigkeit während audiovisueller sexueller Stimulation und/oder intrakavernöser Schwellkörperinjektionstestung ist abhängig von der subjektiven Einschätzung des Untersuchers. Die kontinuierliche Aufzeichnung der Penisrigidität mit Hilfe dem Penis angelegter Meßschleifen ermöglicht eine objektive Dokumentation der Rigidität während der Untersuchung, erlaubt aber keine Aussage über eine Längen-, Durchmesser-oder Volumenzunahme des Penis oder Deviation

more »

... es Gliedes bei Erektion. In Experimenten mit Penismodellen konnten wir demonstrieren, daß ein von uns konzipiertes Sichtprüfsystem ("PRIAPUS") exakt und reproduzierbar berührungsfrei Länge, Durchmesser und Volumen mißt (7). Die erste praktische Anwendung dieses Real-Time-Photogrammetrie-Systems im Vergleich mit der RigiScan®-Aufzeichnung bei 30 Probanden zeigte eine signifikante Korrelation der Rigidität mit der Zunahme des maximalen Penisdurchmessers und des Volumens. Bei 11 Probanden mit einer Rigidität > 70 % wurde die physiologische Kurvatur des erigierten Penis mit 0,05 Grad/cm bestimmt. Bei der Diagnostik der erektilen Dysfunktion mit Schwellkörperinjektionstestung und Inspektion und Palpation des Penis oder Rigiditätsaufzeichnung mit Hilfe dem Penis angelegter Meßschleifen dringt der Untersucher in einen sehr intimen Bereich des Patienten vor. Allein aufgrund der Untersuchungssituation ist mit einer negativen Beeinflussung der aufgezeichneten Erektionsfähigkeit zu rechnen. Die Beurteilung der Penisformveränderung des flakziden und erigierten Penis obliegt der subjektiven Einschätzung des Untersuchers. Das von Akt. Urol. 24 (1993) 95-99 Assessment of erectile capabilities by inspection or palpation during audiovisual sexual Stimulation and/or after intracavernous injection of vasoactive drugs depends upon the examiner's subjective estimation. Penile rigidity recordings with the help of measuring loops attached to the penis objectively document rigidity during examination, but do not provide Information about changes in penile length, diameter volume and curvature. In an experimental set-up with penile models we have already demonstrated that the penile measurements obtained by an image examination System ("PRIAPUS") using photogrammetric tools are exact and reproducible (7). The first practical use of the System with 30 volunteers showed a significant correlation of rigidity and increase of maximum diameter and volume as compared to RigiScan® recordings. In 11 volunteers with a rigidity > 70 % we recorded a physiological curvature of the erect penis of 0.05 degrees/cm. Key words Photogrammetry -Erectile dysfunction -Image examination system Bradley et al. 1985 (2) eingeführte RigiScan®-Gerät ermöglicht durch zwei dem Penis angelegte Meßschleifen neben der Tumeszenzaufzeichnung (Umfangsänderung) auch eine Registrierung der zirkumferenten Penisrigidität, die linear mit der axialen Rigidität und dem intrakavernösen Druck korreliert (4). Trotz akkurater Aufzeichnung nokturner peniler Tumeszenz und Rigidität (NPTR) über mehrere Nächte, kann eine Diskrepanz zwischen diesem Ergebnis und der Erektionsfähigkeit bei entsprechender sexueller Stimulation bestehen (1,3). Aus diesem Grunde führen wir die von Giesbers et al. (5) beschriebene Real-Time-RigiScan®-Aufzeichnung während audiovisueller sexueller Stimulation (AVSS) und intrakavernöser

doi:10.1055/s-2008-1058281 fatcat:jkzw4t2pbfa7fbsvoewbqh6pwq

lang:de

Examining blood microscopic images for leukemia is necessary when expensive equipment for flow cytometry is unavailable. Automated systems can ease the burden on medical experts for performing this examination and may be especially helpful to quickly screen a large number of patients. We present a simple, yet effective classification approach using a ResNeXt convolutional neural network with Squeeze-and-Excitation modules. The approach was evaluated in the C-NMC online challenge and achieves a

more »

... eighted F1-score of 88.91% on the test set. Code is available at https://github.com/jprellberg/isbi2019cancer

arXiv:1906.09020v2 fatcat:bbe3sg2itndxlby47g5mps57ya

Multiple Versions

Objective: Angiogenesis is essential for tumor growth and progression. However, reported data on angiogenic parameters in patients with renal cell carcinoma are contradictory. The objective of this study was to use serum to compare the systemic angiogenic activity in patients with renal cell carcinoma and to determine if pathologic stage and grade correlated to this angiogenesis parameter. Methods: Serum of 28 patients with a newly diagnosed renal cell carcinoma, 28 healthy volunteers and 9

more »

... ents with bladder carcinoma were used for this study. All sera were tested in a 72-hour endothelial cell proliferation assay. In addition the serum concentrations of the angiogenesis stimulators basic ®broblast growth factor (bFGF) and vascular endothelial growth factor (VEGF) were determined using standard ELISA assays. Results: The serum of renal cell carcinoma patients showed a median stimulation of human umbilical vein endothelial cells (HUVEC) of 89.79% (range 58.47±147.95%) and serum of healthy volunteers showed a median stimulation of 95.35% (range 74.64±141.77%) ( p > 0:05). In contrast serum of patients with bladder carcinoma showed a median stimulation of 140.16% (range 64.82±200.16%) ( p 0:024). No correlations of the serum angiogenic activity and tumor stage or grade have been found in renal cell carcinoma patients. Furthermore, no correlations for serum bFGF and VEGF concentrations have been found. Conclusions: Serum angiogenic activity of patients with renal cell carcinoma did not differ signi®cantly from healthy controls, while serum of patients with bladder carcinoma showed a signi®cant increase in endothelial cell stimulation. Furthermore, bFGF and VEGF serum concentrations did not correlate to serum angiogenic activity in patients with renal cell carcinoma. Therefore, the determination of systemic angiogenic parameters, in case of renal cell carcinoma, might not lead to adequate data concerning prognosis or therapeutic effects. #

doi:10.1016/s0302-2838(02)00359-7 pmid:12361902 fatcat:265rmburq5dzxa6senxlkqd6im

Abstract. The low-frequency emissions from a generic 5MW turbine are investigated numerically. In order to regard airborne noise and structure-borne noise simultaneously a process chain was developed. It considers fluid-structure coupling (FSC) of a computational fluid dynamics (CFD) solver and multibody simulations (MBS) solver as well as a Ffowcs Williams-Hawkings (FW-H) acoustic solver. The approach was applied to a generic 5<span

more »

... space">MW turbine to get more insight into the sources and mechanisms of low-frequency emissions from wind turbines. For this purpose simulations with increasing complexity in terms of considered components in the CFD model, degrees of freedom in the structural model and inflow in the CFD model were conducted. Consistent with literature, it has been found that aeroacoustic low-frequency emission is dominated by the blade-passing frequency harmonics. The tower base loads, which excite seismic emission, tend to be dominated by structural eigenfrequencies with increasing complexity of the model. The main source of aeroacoustic emissions is the blade-tower interaction and the contribution of the tower as an acoustic emitter is stronger than the contribution of the rotor. Aerodynamic tower loads also significantly contribute to the external excitation acting on the structure of the wind turbine.

doi:10.5194/wes-2018-51 fatcat:mlfusjre7jd3bixjgdfujhmolm

Open Access

Smooth Normalizing Flows [article]

Preserved Fulltext

Multi-label Classification of Surgical Tools with Convolutional Neural Networks [article]

Preserved Fulltext

Lamarckian Evolution of Convolutional Neural Networks [article]

Preserved Fulltext

Demand-side determinants of rising hospital admissions in Germany: the role of ageing

Preserved Fulltext

Lamarckian Evolution of Convolutional Neural Networks [chapter]

Preserved Fulltext

Determining the refractive index of human hemoglobin solutions by Kramers–Kronig relations with an improved absorption model

Preserved Fulltext

Limited Evaluation Evolutionary Optimization of Large Neural Networks [chapter]

Preserved Fulltext

A cautionary note on computing conditional from unconditional correlations

Preserved Fulltext

Rhombic organization of microvilli domains found in a cell model of the human intestine

Preserved Fulltext

Learned Weight Sharing for Deep Multi-Task Learning by Natural Evolution Strategy and Stochastic Gradient Descent [article]

Preserved Fulltext

New molecular mediators in tumor angiogenesis

Preserved Fulltext

Real-Time-Photogrammetrie in der Urologie

Preserved Fulltext

Acute Lymphoblastic Leukemia Classification from Microscopic Images using Convolutional Neural Networks [article]

Preserved Fulltext

Other Versions

Serum Angiogenic Activity: Diagnostic Relevance in Renal Cell Carcinoma

Preserved Fulltext

Advanced CFD-MBS coupling to assess low-frequency emissions from wind turbines

Preserved Fulltext