The conference Neural Information Processing Systems (NIPS) evolved into the most important conference in the area of machine learning and, in particular, deep learning. The conference with about 8,000 participants enjoys a strong industry participation. Next to the most prominent researchers in machine learning, representatives of the largest big data companies such as Amazon, Microsoft, Google, Facebook, Intel, Tencet and Alibaba attended the event.
The trends at this year's event included interpretability, interdisciplinary application of machine learning, meta-learning and scalable machine learning. This makes a successful participation of the BBDC even more important. With the two presented research papers, "An Empirical Study on The Properties of Random Bases for Kernel Methods" and "SchNet: A continuous-filter convolutional neural network for modeling quantum interactions", at the main conference and two well-attended workshop, "Interpreting, Explaining and Visualizing Deep Learning - Now what ?" and "ML for Materials and Molecules", the machine learning group of TU Berlin made a significant contribution to the overall event. The presented papers address the intersection of an analysis of the intersection of the learning method random features, which received the test-of-time award at this year's conference, and neural networks and about an enhancement of the important convolutional filters for deep learning for continuous spaces, which for example occur in material science.
Already years ago the BBDC-vice-director Klaus-Robert Müller showed vision for future trends when he and his group pioneered in nowadays established subjects such as interpretability of machine learning and the interdisciplinary application of machine learning in the fields of neurotechnology and quantum chemistry. The gathered expertise as well as the created networks in both fields were used to further the research by hosting two successful workshops. Each event showed a large number of attendants and renowned presenters. Among many more, Alex Smola, Kieron Burke, Alán Aspuru-Guzik, Klavs F. Jensen, Risi Kondor, Anatole von Lilienfeld and Oriol Vinyals at the workshop of quantum chemistry as well as Sepp Hochreiter, Honglak Lee, Dhruv Batra, Anh Nguyen, Been Kim and Rich Caruana at the workshop of interpretability presented their newest research. The remarkably productive exchange among the research community, in particular the reunion of research collaborates as well as valuable discussions in and around the workshop, will serve as impulse for future work.