*ANNOUNCEMENT AND CALL FOR CONTRIBUTIONS* ** * The 2019 Conference on Mathematical Theory of Deep Neural Networks (DeepMath 2019) Princeton Club, New York City, Oct 31-Nov 1 2019. Web: https://www.deepmath-conference.com/ ======= Important Dates ======= Submission deadline for 1-page abstracts: June 28, 2019 Notification: TBA. Conference: Oct 31-Nov 1 2019. ======= Confirmed speakers ======= Anima Anandkumar (CalTech), Yasaman Bahri (Google), Minmin Chen (Google), Michael Elad (Technion), Surya Ganguli (Stanford), Tomaso Poggio (MIT), David Schwab (CUNY), Shai Shalev-Shwartz (Hebrew University), Haim Sompolinsky (Hebrew University and Harvard), and Naftali Tishby (Hebrew University). ======= Workshop topic ======= Recent advances in deep neural networks (DNNs), combined with open, easily-accessible implementations, have made DNNs a powerful, versatile method used widely in both machine learning and neuroscience. These advances in practical results, however, have far outpaced a formal understanding of these networks and their training. Recently, long-past-due theoretical results have begun to emerge, shedding light on the properties of large, adaptive, distributed learning architectures. Following the success of the 2018 IAS-Princeton joint symposium on the same topic (https://sites.google.com/site/princetondeepmath/home), the 2019 meeting is more centrally located and broader in scope, but remains focused on rigorous theoretical understanding of deep neural networks. ======= Call for abstracts ======= In addition to these high-profile invited speakers, we invite 1-page non-archival abstract submissions. Abstracts will be reviewed double-blind and presented as posters. To complement the wealth of conferences focused on applications, all submissions for DeepMath 2019 must target theoretical and mechanistic understanding of the underlying properties of neural networks. Insights may come from any discipline and we encourage submissions from researchers working in computer science, engineering, mathematics, neuroscience, physics, psychology, statistics, or related fields. Topics may address any area of deep learning theory, including architectures, computation, expressivity, generalization, optimization, representations, and may apply to any or all network types including fully connected, recurrent, convolutional, randomly connected, or other network topologies. Committee Information: Organizing committee: * Ahmed El Hady * Adam Charles * Mikio Aoi * Andrew Saxe * Joan Bruna * Michael Shvartsman * Sebastian Musslick * Stephen Keeley Advisory committee * Jonathan Cohen * Daniel Lee * Sebastian Seung * Ted Willke Local Committee * NYU: Joan Bruna * Columbia: Ashok Kumar * SUNY Stonybrook: Il Memming Park * Yale: Gal Mishne * City College: David Schwab * IAS: Nadav Cohen *