Ramon.Space (Israel)

The success of deep neural networks (DNNs) is heavily dependent on computational resources. While DNNs are often employed on cloud servers, there is a growing need to operate DNNs on edge devices. Edge devices are typically limited in their computational resources, yet, often multiple edge devices are deployed in the same environment and can reliably communicate with each other. In this work we propose to facilitate the application of DNNs on the edge by allowing multiple users to collaborate during inference to improve their accuracy. Our mechanism, coined edge ensembles, is based on having diverse predictors at each device, which form an ensemble of models during inference. To mitigate the communication overhead, the users share quantized features, and we propose a method for aggregating multiple decisions into a single inference rule. We analyze the latency induced by edge ensembles, showing that its performance improvement comes at the cost of a minor additional delay under common assumptions on the communication network. Our experiments demonstrate that collaborative inference via edge ensembles equipped with compact DNNs substantially improves the accuracy over having each user infer locally, and can outperform using a single centralized DNN larger than all the networks in the ensemble together.

QI2S is a project financed in the frame of the European FP7 funding programme, developing the prototype of a lightweight low-power on-board processing platform that operates with a massively parallel many-core technology based on Ramon Chips ITAR free RC64 architecture. This technology will be able, in the future, of reaching performances of hundreds of GOPS and tens of GFLOPS, although for demonstrational purposes it is now implemented on a downscaled FPGA. The hardware platform is demonstrated for real-time on-board analysis of hyperspectral data. QI2S was developed to ensure parallel computational capabilities and an easily reconfigurable architecture, able to generate on-board products in real-time or near real-time. QI2S is a prototype developed and validated in laboratory with a TRL 4.

On any spatial scale, the species composition of a taxonomic group often departs from a phylogenetically random subset drawn from the pool of species available on a higher scale. Analysis of the uneven representation of related lineages in different assemblages can reveal the action of various forces shaping their diversification. For any assemblage, unequal diversification among lineages can be estimated using diversity skewness, an index of the balance of a phylogenetic tree whose values increase with increasing differences in diversification rates among tree branches. We tested for geographical patterns in the diversity skewness of flea assemblages parasitic on small mammals in 26 distinct geographic localities from the Palaearctic and 15 from the Nearctic. Overall, diversity skewness of the Nearctic flea assemblage was unexpectedly high compared to that of the global flea fauna, whereas that of the Palaearctic did not depart from the expectations of a null model. On a smaller scale, the diversity skewness of local flea assemblages was sometimes lower, sometimes higher, but, in most of the 41 localities, it did not differ significantly from that of random subsets taken from the species pool available on the larger spatial scale (either the world fauna or that of the biogeographical realm, i.e. Palaearctic or Nearctic). More importantly, among Palaearctic assemblages, diversity skewness increased with increasing latitude and/or decreasing mean air temperatures. The different patterns observed in the Palaearctic and Nearctic may be in part due the fact that flea diversification appears to have been more intense in the former than the latter, and to differences between them in relief and glacial history. Temperature-driven speciation rates may well explain the latitudinal gradient in diversity skewness in the Palaearctic. The results illustrate the action of various biogeographical processes in shaping the uneven differentiation of flea lineages on different spatial scales.

Abstract Flexibility in the echolocation call structure of bats can improve their performances, because, in some situations, some signal designs are better than others. Hence, at least some bats should adjust their echolocation calls according to the setting in which they are operating but also to the specific task at hand, that is their behavioral intention. We studied variation in the echolocation calls ofPipistrellus kuhliiemitted during four flight situations that were similar in setting but differed in behavioral context: emergence from a roost, commuting to and from foraging sites, foraging and returning to a roost. Echolocation calls produced byP. kuhliidiffered significantly according to the flight situation. Call types differed most distinctly between foraging and commuting. We also found a high variance in the emergence calls we recorded, perhaps reflecting pre‐ and post‐takeoff calls. Discriminant function analysis on calls emitted while foraging, commuting or returning to the roost classified the calls to the correct group 73.3% of the time. The differences between bats' echolocation calls in different flight situations might indicate an intrinsic change in the bat's behavior. Recognizing these differences could be crucial when using call variables to identify bat species.

Context. The geometry and physical conditions in solar coronal loops are complicated and still not understood well. Recent high-resolution observations obtained with TRACE indicate the existence of sub-resolution transverse structuring not accessible to direct observation. This ingredient has not yet been taken into account in the previous theoretical models used for the study of transversal coronal loop oscillations and of their damping due to resonant conversion of energy. Aims. This study aims to assess the effect of the possibly unresolved internal structure of a coronal loop on the properties of its transverse oscillations and on the efficiency of resonant absorption as a damping mechanism of these oscillations. Methods. The equilibrium configuration of a single coronal loop with internal density structuring is modelled by considering the loop as composed of two very close, parallel, identical coronal slabs in Cartesian geometry. The period of the oscillation and the damping time are computed for the resonantly damped fundamental kink mode. These quantities are then compared to those obtained for two models for a single equivalent slab without internal density structuring. Results. We find that the period and the damping time of a coronal loop with internal density structuring change by less than 15%, when compared to the same oscillatory properties of a single coronal loop with either the same density contrast or a single coronal loop with the same total mass. Conclusions. Therefore the internal density structuring of a coronal loop does not affect its oscillatory properties very much. However, the sub-resolution structuring of a coronal loop with different densities in its components or with different widths could vary these results.

Cestodes are diverse, ubiquitous parasites of vertebrates. They are frequent parasites of small mammals. It is not possible to provide a concise description of their taxonomy and life cycles without omitting many of the complexities, controversies and fascinating novelties of such a diverse group. Therefore, this chapter is introductory and not exhaustive. Its aim is to introduce the reader to the biology of cestodes and to present the basic components of their diversity in rodents, insectivores, lagomorphs and chiropterans as well as the major patterns of cestode life cycles, in which these mammalian groups participate. The appearance of cestodes in their contemporary vertebrate host range comes as a result of multiple independent evolutionary invasions, host shifts and coevolutionary history with their hosts. Appearing well before the origin of mammals, cestodes have not failed to follow their vertebrate hosts wherever they have gone. Not least, whether as intermediate or definitive hosts, small mammals have provided significant opportunities for cestodes to diverge in space and time. Likewise, cestodes have played an integral role in the evolutionary ecology of small mammals.