Uranie 
The high-performance computing assets and the fixed enchancment of each numerical simulation accuracy and the experimental measurements with which they’re confronted, convey a brand new obligatory step to strengthen the credence given to the simulation outcomes: uncertainty quantification. This may have completely different meanings, in accordance with the requested targets (rank uncertainty sources, scale back them, estimate exactly a essential threshold or an optimum working level) and it may request mathematical strategies with better or lesser complexity. This paper introduces the Uranie platform, an Open-source framework which is at present developed on the Various Energies and Atomic Vitality Fee (CEA), within the nuclear power division, with the intention to cope with uncertainty propagation, surrogate fashions, optimisation points, code calibration… This platform advantages from each its dependencies, but additionally from private developments, to supply an environment friendly information dealing with mannequin, a C++ and Python interpreter, superior graphical instruments, a number of parallelisation options… These strategies are very generic and might then be utilized to many sorts of code (as Uranie considers them as black packing containers) so to many fields of physics as properly. On this paper, the instance of thermal trade between a plate-sheet and a fluid is launched to point out how Uranie can be utilized to carry out a wide range of research. The code used to supply the figures of this paper may be present in https://…/uranie together with the sources of the platform. …
Least Absolute Deviation (LAD)
Least absolute deviations (LAD), also called least absolute errors (LAE), least absolute worth (LAV), least absolute residual (LAR), sum of absolute deviations, or the L1 norm situation, is a statistical optimality criterion and the statistical optimization approach that depends on it. Much like the favored least squares approach, it makes an attempt to discover a operate which carefully approximates a set of information. Within the easy case of a set of (x,y) information, the approximation operate is an easy ‘development line’ in two-dimensional Cartesian coordinates. The tactic minimizes the sum of absolute errors (SAE) (the sum of absolutely the values of the vertical ‘residuals’ between factors generated by the operate and corresponding factors within the information). The least absolute deviations estimate additionally arises as the utmost probability estimate if the errors have a Laplace distribution. …
Hierarchical Data Format (HDF)
Hierarchical Information Format (HDF, HDF4, or HDF5) is a set of file codecs and libraries designed to retailer and set up massive quantities of numerical information. Initially developed on the Nationwide Heart for Supercomputing Purposes, it’s supported by the non-profit HDF Group, whose mission is to make sure continued growth of HDF5 applied sciences, and the continued accessibility of information saved in HDF. …
Infinite Latent Feature Selection
Characteristic choice is taking part in an more and more important position with respect to many laptop imaginative and prescient purposes spanning from object recognition to visible object monitoring. Nonetheless, a lot of the current options in characteristic choice should not strong throughout completely different and heterogeneous set of information. On this paper, we tackle this situation proposing a strong probabilistic latent graph-based characteristic choice algorithm that performs the rating step whereas contemplating all of the attainable subsets of options, as paths on a graph, bypassing the combinatorial downside analytically. An interesting attribute of the strategy is that it goals to find an abstraction behind low-level sensory information, that’s, relevancy. Relevancy is modelled as a latent variable in a PLSA-inspired generative course of that permits the investigation of the significance of a characteristic when injected into an arbitrary set of cues. The proposed technique has been examined on ten numerous benchmarks, and in contrast in opposition to eleven cutting-edge characteristic choice strategies. Outcomes present that the proposed strategy attains the best efficiency ranges throughout many various situations and difficulties, thereby confirming its robust robustness whereas setting a brand new cutting-edge in characteristic choice area. …
