BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:Machine Learning on GPUs
DTSTART;VALUE=DATE-TIME:20200421T070000Z
DTEND;VALUE=DATE-TIME:20200421T150000Z
DTSTAMP;VALUE=DATE-TIME:20240704T115516Z
UID:indico-event-2@meetings.aip.de
DESCRIPTION:An overview of activation function types and their application
 s: ReLU\, Sigmoid\, Tanh\n• Cross Entropy: efficient loss beyond MSE\n
 • Softmax - multiclass logistic function\n• Intersection over union me
 tric\n\n• Using PyTorch datasets to organize the data for batch training
 \n• Unet structure. Semantic Image segmentation\n\n• An overview of GP
 U performance in various applications\n• Brief intercomparison of differ
 ent types of accelerators\n• Key programming principles to achieve high 
 GPU performance in Machine Learning\n• Benchmarking CNNs on different GP
 Us: cnn-benchmarks test suite\n\n• An overview of GPU-enabled libraries 
 for feature extraction: torchvision.transform\, Pandas\, OpenCV\, CUBLAS\,
  CUFFT\n\n• Performance limitations of pure Python code\n• Implementin
 g high-performance C++/Fortran/OpenCL extensions to PyTorch with pybind11\
 n\n• Analyzing GPU efficiency for training and inference\, by example of
  VoiceLoop – a neural text-to-speech processor\n\nhttps://meetings.aip.d
 e/event/5/sessions/2/
LOCATION:AIP SH/Lecture Hall
URL:https://meetings.aip.de/event/5/sessions/2/
END:VEVENT
END:VCALENDAR