Technical University of Berlin

About: Technical University of Berlin is a education organization based out in Berlin, Germany. It is known for research contribution in the topics: Laser & Catalysis. The organization has 27292 authors who have published 59342 publications receiving 1414623 citations. The organization is also known as: Technische Universität Berlin & TU Berlin.

Transiting exoplanets from the CoRoT space mission - II. CoRoT-Exo-2b: a transiting planet around an active G star

Abstract: Context. The CoRoT mission, a pioneer in exoplanet searches from space, has completed its first 150 days of continuous observations of ∼12 000 stars in the galactic plane. An analysis of the raw data identifies the most promising candidates and triggers the ground-based follow-up. Aims. We report on the discovery of the transiting planet CoRoT-Exo-2b, with a period of 1.743 days, and characterize its main parameters. Methods. We filter the CoRoT raw light curve of cosmic impacts, orbital residuals, and low frequency signals from the star. The folded light curve of 78 transits is fitted to a model to obtain the main parameters. Radial velocity data obtained with the SOPHIE, CORALIE and HARPS spectrographs are combined to characterize the system. The 2.5 min binned phase-folded light curve is affected by the effect of sucessive occultations of stellar active regions by the planet, and the dispersion in the out of transit part reaches a level of 1.09 × 10 −4 in flux units. Results. We derive a radius for the planet of 1.465 ± 0.029 RJup and a mass of 3.31 ± 0.16 MJup, corresponding to a density of 1.31 ± 0.04 g/cm 3 . The large radius of CoRoT-Exo-2b cannot be explained by current models of evolution of irradiated planets.

Clustered Federated Learning: Model-Agnostic Distributed Multitask Optimization Under Privacy Constraints

Abstract: Federated learning (FL) is currently the most widely adopted framework for collaborative training of (deep) machine learning models under privacy constraints. Albeit its popularity, it has been observed that FL yields suboptimal results if the local clients’ data distributions diverge. To address this issue, we present clustered FL (CFL), a novel federated multitask learning (FMTL) framework, which exploits geometric properties of the FL loss surface to group the client population into clusters with jointly trainable data distributions. In contrast to existing FMTL approaches, CFL does not require any modifications to the FL communication protocol to be made, is applicable to general nonconvex objectives (in particular, deep neural networks), does not require the number of clusters to be known a priori , and comes with strong mathematical guarantees on the clustering quality. CFL is flexible enough to handle client populations that vary over time and can be implemented in a privacy-preserving way. As clustering is only performed after FL has converged to a stationary point, CFL can be viewed as a postprocessing method that will always achieve greater or equal performance than conventional FL by allowing clients to arrive at more specialized models. We verify our theoretical analysis in experiments with deep convolutional and recurrent neural networks on commonly used FL data sets.