-
A Deep Neural Network -- Mechanistic Hybrid Model to Predict Pharmacokinetics in Rat
Authors:
Florian Führer,
Andrea Gruber,
Holger Diedam,
Andreas H. Göller,
Stephan Menz,
Sebastian Schneckener
Abstract:
An important aspect in the development of small molecules as drugs or agro-chemicals is their systemic availability after intravenous and oral administration. The prediction of the systemic availability from the chemical structure of a potential candidate is highly desirable, as it allows to focus the drug or agrochemical development on compounds with a favorable kinetic profile. However, such pre…
▽ More
An important aspect in the development of small molecules as drugs or agro-chemicals is their systemic availability after intravenous and oral administration. The prediction of the systemic availability from the chemical structure of a potential candidate is highly desirable, as it allows to focus the drug or agrochemical development on compounds with a favorable kinetic profile. However, such pre-dictions are challenging as the availability is the result of the complex interplay between molecular properties, biology and physiology and training data is rare. In this work we improve the hybrid model developed earlier [1]. We reduce the median fold change error for the total oral exposure from 2.85 to 2.35 and for intravenous administration from 1.95 to 1.62. This is achieved by training on a larger data set, improving the neural network architecture as well as the parametrization of mechanistic model. Further, we extend our approach to predict additional endpoints and to handle different covariates, like sex and dosage form. In contrast to a pure machine learning model, our model is able to predict new end points on which it has not been trained. We demonstrate this feature by predicting the exposure over the first 24h, while the model has only been trained on the total exposure.
△ Less
Submitted 2 January, 2024; v1 submitted 13 October, 2023;
originally announced October 2023.
-
Towards online triggering for the radio detection of air showers using deep neural networks
Authors:
Florian Führer,
Tom Charnock,
Anne Zilles,
Matias Tueros
Abstract:
The detection of air-shower events via radio signals requires to develop a trigger algorithm for a clean discrimination between signal and background events in order to reduce the data stream coming from false triggers. In this contribution we will describe an approach to trigger air-shower events on a single-antenna level as well as performing an online reconstruction of the shower parameters usi…
▽ More
The detection of air-shower events via radio signals requires to develop a trigger algorithm for a clean discrimination between signal and background events in order to reduce the data stream coming from false triggers. In this contribution we will describe an approach to trigger air-shower events on a single-antenna level as well as performing an online reconstruction of the shower parameters using neural networks.
△ Less
Submitted 7 December, 2018; v1 submitted 6 September, 2018;
originally announced September 2018.
-
Renormalizing a Viscous Fluid Model for Large Scale Structure Formation
Authors:
Florian Führer,
Gerasimos Rigopoulos
Abstract:
Using the Stochastic Adhesion Model (SAM) as a simple toy model for cosmic structure formation, we study renormalization and the removal of the cutoff dependence from loop integrals in perturbative calculations. SAM shares the same symmetry with the full system of continuity+Euler equations and includes a viscosity term and a stochastic noise term, similar to the effective theories recently put fo…
▽ More
Using the Stochastic Adhesion Model (SAM) as a simple toy model for cosmic structure formation, we study renormalization and the removal of the cutoff dependence from loop integrals in perturbative calculations. SAM shares the same symmetry with the full system of continuity+Euler equations and includes a viscosity term and a stochastic noise term, similar to the effective theories recently put forward to model CDM clustering. We show in this context that if the viscosity and noise terms are treated as perturbative corrections to the standard eulerian perturbation theory, they are necessarily non-local in time. To ensure Galilean Invariance higher order vertices related to the viscosity and the noise must then be added and we explicitly show at one-loop that these terms act as counter terms for vertex diagrams. The Ward Identities ensure that the non-local-in-time theory can be renormalized consistently. Another possibility is to include the viscosity in the linear propagator, resulting in exponential damping at high wavenumber. The resulting local-in-time theory is then renormalizable to one loop, requiring less free parameters for its renormalization.
△ Less
Submitted 26 January, 2016; v1 submitted 10 September, 2015;
originally announced September 2015.
-
Backreaction in Growing Neutrino Quintessence
Authors:
Florian Führer,
Christof Wetterich
Abstract:
We investigate the cosmological effects of neutrino lumps in Growing Neutrino Quintessence. The strongly non-linear effects are resolved by means of numerical N-body simulations which include relativistic particles, non-linear scalar field equations and backreaction effects. For the investigated models with a constant coupling between the scalar field and the neutrinos the backreaction effects are…
▽ More
We investigate the cosmological effects of neutrino lumps in Growing Neutrino Quintessence. The strongly non-linear effects are resolved by means of numerical N-body simulations which include relativistic particles, non-linear scalar field equations and backreaction effects. For the investigated models with a constant coupling between the scalar field and the neutrinos the backreaction effects are so strong that a realistic cosmology is hard to realize. This points towards the necessity of a field dependent coupling in Growing Neutrino Quintessence. In this case realistic models of dynamical Dark Energy exist which are testable by the observation or non-observation of large neutrino lumps.
△ Less
Submitted 27 March, 2015;
originally announced March 2015.
-
Higher-order massive neutrino perturbations in large-scale structure
Authors:
Florian Führer,
Yvonne Y. Y. Wong
Abstract:
We develop a higher-order perturbation theory for large-scale structure formation involving a free-streaming hot or warm dark matter species. We focus on the case of mixed cold dark matter and massive neutrinos, although our approach is applicable also to a single warm dark matter species. In order to capture the suppressed growth of neutrino density perturbations on small scales, we account for…
▽ More
We develop a higher-order perturbation theory for large-scale structure formation involving a free-streaming hot or warm dark matter species. We focus on the case of mixed cold dark matter and massive neutrinos, although our approach is applicable also to a single warm dark matter species. In order to capture the suppressed growth of neutrino density perturbations on small scales, we account for the full momentum dependence of the phase space distribution using the Vlaslov equation, and derive from it a formal closed-form nonlinear equation for the neutrino density. Using a systematic perturbative expansion of this equation we compute high-order corrections to the neutrino density contrast without the explicit need to track the perturbed neutrino momentum distribution. We calculate the leading-order total matter bispectrum for several neutrino masses. Using our result as a benchmark, we test the accuracy of the fluid approximation and a linear approximation used in perturbative and N-body analyses, as well as a new hybrid approach that combines the exact linear evolution with the nonlinear structure of the fluid equations. Aiming at $\lesssim1%$ accuracy, we find that the total matter bispectrum with a low neutrino mass m=0.046 eV can be reproduced by all but the fluid approximation, while for larger neutrino masses m=0.46-0.93 eV only the hybrid approach has the desired accuracy on a large range of scales. This result serves as a cautionary note that approximate nonlinear models of neutrino clustering that reproduce the gross features of some observables may not suffice for precision calculations, nor are they guaranteed to apply to other observables. All of the approximation schemes fail to reproduce the bispectrum of the neutrino density perturbations at better than 20% accuracy across all scales, indicating that an exact treatment of nonlinear neutrino perturbations is necessary.
△ Less
Submitted 3 March, 2015; v1 submitted 8 December, 2014;
originally announced December 2014.
-
Nonlinear growing neutrino cosmology
Authors:
Youness Ayaita,
Marco Baldi,
Florian Führer,
Ewald Puchwein,
Christof Wetterich
Abstract:
The energy scale of Dark Energy, $\sim 2 \times 10^{-3}$ eV, is a long way off compared to all known fundamental scales - except for the neutrino masses. If Dark Energy is dynamical and couples to neutrinos, this is no longer a coincidence. The time at which Dark Energy starts to behave as an effective cosmological constant can be linked to the time at which the cosmic neutrinos become nonrelativi…
▽ More
The energy scale of Dark Energy, $\sim 2 \times 10^{-3}$ eV, is a long way off compared to all known fundamental scales - except for the neutrino masses. If Dark Energy is dynamical and couples to neutrinos, this is no longer a coincidence. The time at which Dark Energy starts to behave as an effective cosmological constant can be linked to the time at which the cosmic neutrinos become nonrelativistic. This naturally places the onset of the Universe's accelerated expansion in recent cosmic history, addressing the why-now problem of Dark Energy. We show that these mechanisms indeed work in the Growing Neutrino Quintessence model - even if the fully nonlinear structure formation and backreaction are taken into account, which were previously suspected of spoiling the cosmological evolution. The attractive force between neutrinos arising from their coupling to Dark Energy grows as large as $10^6$ times the gravitational strength. This induces very rapid dynamics of neutrino fluctuations which are nonlinear at redshift $z \approx 2$. Nevertheless, a nonlinear stabilization phenomenon ensures only mildly nonlinear oscillating neutrino overdensities with a large-scale gravitational potential substantially smaller than that of cold dark matter perturbations. Depending on model parameters, the signals of large-scale neutrino lumps may render the cosmic neutrino background observable.
△ Less
Submitted 31 July, 2014;
originally announced July 2014.
-
Mirrored strain-balanced quantum well concentrator cells in the radiative limit
Authors:
J. P. Connolly,
M. F. Führer,
D. C. Johnson,
I. M. Ballard,
K. W. J. Barnham,
M. Mazzer,
T. N. D Tibbits,
J. S. Roberts,
G. Hill,
C. Calder
Abstract:
Strain-balanced Quantum Well Solar Cells (SB- QWSCs) are radiatively dominated at concentrator current levels. Incorporating back surface mirrors on the back of the doped substrate leads to a reduction in this radiative recombination current while the non-radiative dark current at lower bias remains unchanged. This work extends a previous model of radiative dark currents to cells where the substra…
▽ More
Strain-balanced Quantum Well Solar Cells (SB- QWSCs) are radiatively dominated at concentrator current levels. Incorporating back surface mirrors on the back of the doped substrate leads to a reduction in this radiative recombination current while the non-radiative dark current at lower bias remains unchanged. This work extends a previous model of radiative dark currents to cells where the substrate is not a perfectly absorbing sink for emitted luminescence by including multiple passes of photons emitted by radiative recombination. The luminescence in such mirrored structures is reduced and the radiative dark current lower. The structure is suited to concentrator applications because it corresponds to the high efficiency limit where losses due to light re-emission are limited to the solid angle for absorption of solar radiation.
△ Less
Submitted 9 June, 2010; v1 submitted 3 June, 2010;
originally announced June 2010.
