No abstract

The impact of the maximally possible batch size (for the better runtime) on performance of graphic processing units (GPU) and tensor processing units (TPU) during training and inference phases is investigated. The numerous runs of the selected deep neural network (DNN) were performed on the standard MNIST and Fashion-MNIST datasets. The significant speedup was obtained even for extremely low-scale usage of Google TPUv2 units (8 cores only) in comparison to the quite powerful GPU NVIDIA Tesla K80 card with the speedup up to 10x for training stage (without taking into account the overheads) and speedup up to 2x for prediction stage (with and without taking into account overheads). The precise speedup values depend on the utilization level of TPUv2 units and increase with the increase of the data volume under processing, but for the datasets used in this work (MNIST and Fashion-MNIST with images of sizes 28x28) the speedup was observed for batch sizes >512 images for training phase and >40 000 images for prediction phase. It should be noted that these results were obtained without detriment to the prediction accuracy and loss that were equal for both GPU and TPU runs up to the 3rd significant digit for MNIST dataset, and up to the 2nd significant digit for Fashion-MNIST dataset.

-The paper concerns the hype idea of "Citizen Science" and the related paradigm shift: to go from the passive "volunteer computing" to other volunteer actions like "volunteer measurements" under guidance of scientists. They can be carried out by ordinary people with standard computing gadgets (smartphone, tablet, etc.) and the various standard sensors in them. Here the special attention is paid to the system of volunteer scientific measurements to study air showers caused by cosmic rays. The technical implementation is based on integration of data about registered night flashes (by radiometric software) in shielded camera chip, synchronized time and GPS-data in ordinary gadgets: to identify night "air showers" of elementary particles; to analyze the frequency and to map the distribution of "air showers" in the densely populated cities. The project currently includes the students of the National Technical University of Ukraine "KPI", which are compactly located in Kyiv city and contribute their volunteer measurements. The technology would be very effective for other applications also, especially if it will be automated (e.g., on the basis of XtremWeb or/and BOINC technologies for distributed computing) and used in some small area with many volunteers, e.g. in local communities (Corporative/Community Crowd Computing).

No abstract

No abstract