Tobi Delbruck

Tobi Delbruck
(Co-Director of Sensors Group)

Biography

Tobi Delbruck  (IEEE M’99–SM’06–F’13) received a B.Sc. degree in physics from University of California in 1986 and a Ph.D. degree from Caltech in 1993 in the inaugural class of the Computation and Neural Systems program founded by John Hopfield, as a student of Christof Koch, David van Essen and Carver Mead. Currently he is a Professor of Physics and Electrical Engineering at ETH Zurich in the Institute of Neuroinformatics, University of Zurich and ETH Zurich, Switzerland, where he has been since 1998.  He co-organizes the Telluride Neuromorphic Engineering workshop and has organized live demonstration sessions at ISCAS, NeurIPS, and AICAS and two Confession Sessions at ISCAS. Delbruck is past Chair of the IEEE CAS Sensory Systems Technical Committee. He worked on electronic imaging at Arithmos, Synaptics, National Semiconductor, and Foveon and has co-founded 3 companies inilabs, insightness, and inivation. He invented the adaptive photoreceptor circuit. The MOS pseudo resistor used in it is a key part of the most cited JSSC paper of the 2005-2015 decade in the neural-amplifier paper from R. Harrison. He also invented the “bump” circuit and developed open-source ultra wide dynamic range digitally programmable bias current generators used in many neuromorphic chips. His IEEE J. Solid State Circuits paper on the dynamic vision sensor silicon retina event camera is the 4th most cited in the 2005-2015 decade. These event camera developments inspired the Sensors Group's recent work on activity-driven AI hardware accelerators, e.g. NullHop and DeltaRNN, which are among the first to exploit neuromorphic activation sparsity for saving time and energy like spiking neural networks, but in a way that is much more compatible with using DRAM memory for cost-efficient scaling to large deep networks. Over the past 5 years he has been working towards using these hardware AI circuits for adaptive nonlinear robotic control. His papers have been awarded 13 IEEE awards and in 2013 was named a Fellow of the IEEE Circuits and Systems Society for his work on neuromorphic sensors and processing.  He likes to read storybooks, play tennis, and sometimes tries card magic on unwary subjects.

Work History

Starting around 2003, I worked with Patrick Lichtsteiner on making a high-quality address-event silicon retina, starting from the developments of the late Jorg Kramer (RIP). This work started from the CAVIAR project, the grandparent of all modern event cameras and spike based neuromorphic perception systems.

From 2003-2007, I worked on the adaptive building intelligence project, which was an early attempt to use machine learning for happier building occupants and which led to the concept of a PC presence sensor.

From 2001-2002 I worked with Sam Zahnd on the development of a vision-chip-augmented occupancy detector in collaboration with a Swiss company (HTS) that makes PIR (passive infrared) occupancy detectors.

From 2000-2002 , I took a detour from aVLSI and concentrated nearly a lot of my time on development of the luminous tactile floor for the exhibit “Ada: Intelligent Space” that INI developed for the Swiss Expo.02. This floor acted as the skin of the space, and was the primary interaction medium for the visitors. The Ada exhibit finished with more than 550,000 visitors. It was great to work with Rodney Douglas and Adrian Whatley on the development of Ada's floor.

From 1999-2001 we (Shih-Chii Liu and other hardware group members) developed the Physiologist’s Friend Chip, a self-contained, user-friendly device that models retinal and cortical spiking neurons. This chip is useful as a model animal for physiologists, and for lecturers who wish to demonstrate cell responses. We've sold a number of these devices to physiology labs around the world.

During the period 1995-1998, while consulting in Silicon Valley at several companies; (Arithmos--bought by STMicro, Synaptics, National Semiconductor , and Foveon). I was mainly working on CMOS imager technology. At Synaptics, on a project that led to the formation of Foveon, I worked with Carver Mead on engineering of a new type of; ”pulsed-bipolar” imager, which, though it had some really interesting and unique properties, didn't meet the needs of the marketplace.

After finishing my PhD with Carver I did a year of postdoc in 1994, and spent a few months at Oxford working with Misha Mahowald and Rodney Douglas on a next generation silicon retina (none of these chips worked at all).

Before finishing my PhD during the period 1987-1993 n Carver Mead's Physics of Computation lab at Caltech, I also spent a year in David Van Essen’s lab looking for Charlie Anderson's cortical “shifter” circuits in monkey visual physiology experiments. (We found some intriguing results, but nothing definitive.) Before that, I spent about a year working with Christof Koch on mapping the Marr-Poggio stereopsis algorithm onto a Hopfield network and getting it to show the hysteretic behavior observed in stereo fusion.