CANDLE/Supervisor: a workflow framework for machine learning applied to cancer research.

BACKGROUND: Current multi-petaflop supercomputers are powerful systems, but present challenges when faced with problems requiring large machine learning workflows. Complex algorithms running at system scale, often with different patterns that require disparate software packages and complex data flows cause difficulties in assembling and managing large experiments on these machines. RESULTS: This paper presents a workflow system that makes progress on scaling machine learning ensembles, specifically in this first release, ensembles of deep neural networks that address problems in cancer research across the atomistic, molecular and population scales. The initial release of the application framework that we call CANDLE/Supervisor addresses the problem of hyper-parameter exploration of deep neural networks. CONCLUSIONS: Initial results demonstrating CANDLE on DOE systems at ORNL, ANL and NERSC (Titan, Theta and Cori, respectively) demonstrate both scaling and multi-platform execution.

Validation of ICON-MIGHTI Thermospheric Wind Observations: 1. Nighttime Red-Line Ground-Based Fabry-Perot Interferometers.

Observations of the nighttime thermospheric wind from two ground-based Fabry-Perot Interferometers are compared to the level 2.1 and 2.2 data products from the Michelson Interferometer Global High-resolution Thermospheric Imaging (MIGHTI) onboard National Aeronautics and Space Administration's Ionospheric Connection Explorer to assess and validate the methodology used to generate measurements of neutral thermospheric winds observed by MIGHTI. We find generally good agreement between observations approximately coincident in space and time with mean differences less than 11 m/s in magnitude and standard deviations of about 20-35 m/s. These results indicate that the independent calculations of the zero-wind reference used by the different instruments do not contain strong systematic or physical biases, even though the observations were acquired during solar minimum conditions when the measured airglow intensity is weak. We argue that the slight differences in the estimated wind quantities between the two instrument types can be attributed to gradients in the airglow and thermospheric wind fields and the differing viewing geometries used by the instruments.