Search for: All records

Abstract Modifying turbine blade pitch, generator torque, and nacelle direction (yaw) are conventional approaches for enhancing energy output and alleviating structural loads. However, the efficacy of such methods is challenged by the lag in adjusting such settings after atmospheric variations are detected. Without reliable short-term wind forecasting tools, current practice, which mostly relies on data collected at or just behind turbines, can result in sub-optimal performance. Data-assimilation strategies can achieve real-time wind forecasting capabilities by correcting model-based predictions of the incoming wind using various field measurements. In this paper, we revisit the development of a class of prior models for real-time estimation via Kalman filtering algorithms that track atmospheric variations using ground-level pressure sensors. This class of models is given by the stochastically forced linearized Navier-Stokes equations around the three-dimensional waked velocity profile defined by a curled wake model. The stochastic input to these models is devised using convex optimization to achieve statistical consistency with high-fidelity large-eddy simulations. We demonstrate the ability of such models in reproducing the second-order statistical signatures of the turbulent velocity field. In support of assimilating ground-level pressure measurements with the predictions of said models, we also highlight the significance of the wall-normal dimension in enhancing two-point correlations of the pressure field between the ground and the computational domain. 

ABSTRACT Symbiotic marine bacteria that are transmitted through the environment are susceptible to abiotic factors (salinity, temperature, physical barriers) that can influence their ability to colonize their specific hosts. Given that many symbioses are driven by host specificity, environmentally transmitted symbionts are more susceptible to extrinsic factors depending on conditions over space and time. In order to determine whether the population structure of environmentally transmitted symbionts reflects host specificity or biogeography, we analysed the genetic structure ofSepiola atlantica(Cephalopoda: Sepiolidae) and theirVibriosymbionts (V. fischeriandV. logei) in four Galician Rías (Spain). This geographical location is characterized by a jagged coastline with a deep‐sea entrance into the land, ideal for testing whether such population barriers exist due to genetic isolation. We used haplotype estimates combined with nested clade analysis to determine the genetic relatedness for bothS. atlanticaandVibriobacteria. Analyses of molecular variance (AMOVA) were used to estimate variation within and between populations for both host and symbiont genetic data. Our analyses reveal a low percentage of variation among and between host populations, suggesting that these populations are panmictic. In contrast,Vibriosymbiont populations show certain degree of genetic structure, demonstrating that the hydrology of the rías is driving bacterial distribution (and not host specificity). Thus, for environmentally transmitted symbioses such as the sepiolid squid‐Vibrioassociation, abiotic factors can be a major selective force for determining population structure for one of the partners. 

We provide an information-theoretic analy- sis of the generalization ability of Gibbs- based transfer learning algorithms by focus- ing on two popular empirical risk minimiza- tion (ERM) approaches for transfer learning, α-weighted-ERM and two-stage-ERM. Our key result is an exact characterization of the generalization behavior using the conditional symmetrized Kullback-Leibler (KL) informa- tion between the output hypothesis and the target training samples given the source train- ing samples. Our results can also be applied to provide novel distribution-free generaliza- tion error upper bounds on these two afore- mentioned Gibbs algorithms. Our approach is versatile, as it also characterizes the gener- alization errors and excess risks of these two Gibbs algorithms in the asymptotic regime, where they converge to the α-weighted-ERM and two-stage-ERM, respectively. Based on our theoretical results, we show that the ben- efits of transfer learning can be viewed as a bias-variance trade-off, with the bias induced by the source distribution and the variance induced by the lack of target samples. We believe this viewpoint can guide the choice of transfer learning algorithms in practice. 

Various approaches have been developed to upper bound the generalization error of a supervised learning algorithm. However, existing bounds are often loose and lack of guarantees. As a result, they may fail to characterize the exact generalization ability of a learning algorithm.Our main contribution is an exact characterization of the expected generalization error of the well-known Gibbs algorithm (a.k.a. Gibbs posterior) using symmetrized KL information between the input training samples and the output hypothesis. Our result can be applied to tighten existing expected generalization error and PAC-Bayesian bounds. Our approach is versatile, as it also characterizes the generalization error of the Gibbs algorithm with data-dependent regularizer and that of the Gibbs algorithm in the asymptotic regime, where it converges to the empirical risk minimization algorithm. Of particular relevance, our results highlight the role the symmetrized KL information plays in controlling the generalization error of the Gibbs algorithm. 

Search for a new pseudoscalar a-boson decaying to muons in events with additional top quark pairs.