- Tuesday, 21 February 2017, 2:30PM -- DC 1302
- Computer Science Seminar -- Computer Science
- Speaker: Alexander Wong, Systems Design Engineering, University of Waterloo
Title: "Deep Learning with Darwin: Evolutionary Synthesis of Operational Deep Intelligence"
Abstract: Deep learning has given rise to a major revolution in the field of artificial intelligence (AI). A major challenge with the democratization and proliferation of deep learning as commodity AI for all is the sheer complexity of current deep neural networks, making them ill-suited for operational use in a large number of scenarios. Taking inspiration from biological evolution, this talk explores the idea of "Can deep neural networks evolve naturally over successive generations into highly efficient deep neural networks?". Recent findings will be presented that support such an evolutionary synthesis paradigm for achieving operational deep intelligence across a wide variety of scenarios.
Dr. Alexander Wong is the Canada Research Chair in Medical Imaging Systems, co-director of the Vision and Image Processing Research Group, and an associate professor in the Department of Systems Design Engineering at the University of Waterloo. He has published over 300 refereed journal and conference papers, as well as patents, in various fields such as computational imaging, artificial intelligence, computer vision, and multimedia systems. In the area of artificial intelligence, his focus is on operational deep intelligence (a pioneer in evolutionary deep intelligence, discovery radiomics, and random deep intelligence via deep-structured fully-connected graphical models).
- Friday, 24 February 2017, 1:00PM -- MC 6460
- PhD Defence -- Applied Mathematics
- Speaker: Subasha Wickramarachchi, Department of Applied Mathematics
Title: "The hydrodynamics of two-dimensional oscillating flows over ripples: the effects of asymmetries in ripple shape and currents"
Abstract: The research presented in this thesis involved consideration of three increasingly asymmetric oscillatory flows over two ripple profiles, one of which is asymmetric and represents a small deviation from a symmetric ripple. The asymmetric flow forcings are chosen so as to have an identical excursion distance. For each of these cases, the vorticity, velocity, and shear stress fields are analyzed for the first 40 cycles. The Reynolds number is mostly limited to 1250, with a few cases also run at Re=5000.
During the first three cycles, it can be observed that two vortices of opposite vorticity are generated during each cycle, one each during the forward and backward flows. These vortices pair up and trace a path that extends higher up as the asymmetry of the flow or ripple shape is amplified. This effect is a consequence of the increasing difference between the vorticities of the individual vortices in each vortex pair. After about three periods, the motion becomes more and more chaotic due to the interaction of the newly generated vortices with the older ones.
During the quasi-steady state, which occurs after about 25 cycles, the vortices generated remain close to the ripple and dissipate before reaching much higher for all cases except the all-symmetric case. For the asymmetric ripple and the asymmetric forcings, a consistent negative mean flow can be observed, which tends to attain a near-steady speed after about 35 cycles. During the quasi-steady state, vortex interactions with the ripple contribute a substantial amount of momentum flux to the system, as do the strong shear layers that can be seen to form at a short distance on either side of the ripple crest. The newly generated vortex pairs exhibit initial trajectories in the direction of the mean flow and are found to trace the transfer of momentum from near the ripple to area of mean current.
- Friday, 24 February 2017, 1:00PM -- DC 1331
- ISS4E Seminar -- Computer Science
- Speaker: Florian Kerschbaum, David R. Cheriton School of Computer Science
Title: "Privacy in the Smart Grid / Road Toll Collection"
We present a cryptographic approach to privacy in the Smart Grid / Road Toll Collection. All data is encrypted before transmitting to the service provider and all computations are done on encrypted. We ensure multi-lateral security by using zero-knowledge proofs and controlled observation to ensure the integrity of computation for correct billing. This talk will briefly present protocols for billing, aggregation and general computations in the smart grid and billing in the road toll collection.
Florian is an associate professor in the David R. Cheriton School of Computer Science at the University of Waterloo (since 2017). Before he has worked as a chief research expert at SAP in Karlsruhe (2005 – 2016) and as a software architect at Arxan Technologies in San Francisco (2002 – 2004). His research interests are searching over and generally computation on encrypted data. His work has been applied in the real world to databases, supply chain management and RFID tracking. He holds a Ph.D. in computer science from the Karlsruhe Institute of Technology (2010) and a master's degree from Purdue University (2001).
- Friday, 24 February 2017, 1:30PM *** CANCELLED ***
Number Theory Seminar Seminar -- Pure Mathematics
- Daniel Bertrand, Universite Pierre et Marie Curie, Paris
""Pell equations over polynomial rings and generalized jacobians""
- Friday, 24 February 2017, 2:30PM -- MC 5413
- Geometry & Topology Seminar Seminar -- Pure Mathematics
- Speaker: Susama Agarwala, United States Naval Academy
Title: "“Positive Grassmannians and the Geometry of Wilson Loop diagrams”"
Abstract: In this talk, I introduce a family of diagrams, called Wilson loop diagrams, of physical significance in SYM N=4 theory. There is strong evidence that these diagrams should define a submanifold of the positive Grassmannian Gr≥0(k,n), with k and n determined by the diagram. I present a map from these diagrams to the positive Grasmannians, and explore the geometry of the Wilson Loop diagrams.