From December 3-8, 2018, the Palais des Congrès de Montréal in Canada hosted the Conference on Neural Information Processing Systems (NeurIPS, formerly known as NIPS), one of the world's leading scientific events in the field of Artificial Intelligence and Machine Learning. As part of the conference a workshop titled "Compact Deep Neural Network Representation with Industrial Applications (CDNNRIA)" was co-organized by JOANNEUM RESEARCH.
This workshop aimed to bring together researchers, educators, practitioners who are interested in techniques as well as applications of making compact and efficient neural network representations. One main theme of the workshop discussion was to build up consensus in this rapidly developed field, and in particular, to establish close connection between researchers in Machine Learning community and engineers in industry. The workshop assembled more than 200 participants from academia and industry working on techniques and applications of efficient neural network-based learning.
The workshop hosted invited talks with speakers from MIT, DeepMind, Intel, Qualcomm, NVIDIA and University of Amsterdam, oral and poster presentations and a panel discussion. Among the topics discussed was the closer integration between neural network optimization and target hardware, the question of pruning existing networks vs. training compacter networks from scratch, and the necessity for benchmarking of network compression methods. In order to follow up these topics, which are very relevant for the practical use of neural networks in services such as those used in MARCONI, workshops at ICML and NeurIPS 2019 are planned.