Computer Vision News - February 2021

33 Computational Aspects of Deep Learning Workshop at ICPR2020 was organized with the goal of bringing the focus of the community on computational challenges, and as an opportunity to present solutions and research works in this direction. This first edition has focused on the development of optimized architectures , particularly on highly scalable systems, tackling the training on large-scale datasets, the design of novel architectures and operators, and the training on multi- node systems such as HPC clusters. The workshop started in the morning with an invited tutorial, organized by Adam Henryk Grzywaczewski , on the parallel training of Neural Networks . Adamshowed, througha critical reviewof the literature and practical examples, how to deal with the challenges of very large batch training, a setting that is very frequent in HPC and highly distributed Deep Learning. The second part of the day has instead been dedicated to the presentation of ten research works , which focused on optimization and computational challenges. Beyond research papers, CADL also featured two exciting invited talks in the middle of the program. The first talk was given by Tom Gibbs , Vice President of Business Development at NVIDIA , who has been recently awarded the ACM Gordon Bell prize this year. He