December 18-20, 2018, Hannover, Germany

Interdisciplinary Foundations for Non-Digital Computing

Performance growth of digital microchips is flattening out and energy consumption is becoming a serious issue. This impasse has re-invigorated two lines of investigation that have been explored for decades without yet leading to decisive breakthroughs. The first is learning from the brain with its amazing intelligence-per-watt ratio. The second is the exploration of unconventional physical substrates and nonlinear phenomena.

 

Our conference has two aims: to explore technological options arising from novel computational architectures in unconventional substrates; and to work towards a general, rigorous theory of computing in non-digital, "brain-like" physical substrates.

 

This is a profoundly interdisciplinary enterprise. For the duration of three full days, we want to bring together researchers from neuroscience, mathematics, machine learning, computer engineering, physics and material sciences in a single-track "think tank" setting with eye-opening plenary and keynote lectures, and carefully selected oral presentations, broadened by comprehensive poster sessions.

 

The venue, the Castle of Herrenhausen, a heritage of the Kings of Hannover, was transformed into an award-winning center for scientific events. The Herrenhausen Gardens stretch across hectares of classical French gardening. This unique setting and generous break times will make it easy and natural to engage in a stimulating scientific exchange.

 

Our conference materializes as a Herrenhäuser Conference, a framework offered by the Volkswagen Foundation for hosting events with a particularly interdisciplinary and open-minded character.

 

The event is generously supported by the Volkswagen Foundation.
Oral presenters will receive travel grants.


Download Poster

Invited speakers

Kwabena Boahen

Keynote for Session 3: Neuromorphic Hardware

Professor, Stanford University, USA. Some research themes: neuromorphic engineering, silicon retina, mixed analog-digital platforms, computing with unreliable nanoscale devices, novel computing paradigms.

https://web.stanford.edu/group/brainsinsilicon/boahen.html

Joanna J. Bryson

Plenary Lecture 3: Concerning Ethical and Societal Implications

Professor, University of Bath, UK, and Princeton University, USA. Some research themes: theories of intelligence, AI models of cognitive systems, evolutionary and learning dynamics, action selection, human culture, primate behaviour, robot and AI ethics.

http://www.cs.bath.ac.uk/~jjb/

Chris Eliasmith

Plenary Lecture 2: Concerning Computing Brains

Professor, University of Waterloo, Canada; Director, Centre for Theoretical Neuroscience, Waterloo; Research Chair in Theoretical Neuroscience, Royal Society of Canada. Some research themes: theoretical and computational neuroscience, large-scale simulations of cognitive neural systems, philosophy of mind and language, philosophy of science.

http://arts.uwaterloo.ca/~celiasmi/

Edward A. Lee

Keynote for Session 2: Theoretical Concepts and Mathematical Foundations

Robert S. Pepper Distinguished Professor, UC Berkeley, USA. Some research themes: models of computation with time and concurrency, architectures for real-time and distributed computing, fault tolerance, sensor networks, blending computing with continuous dynamics and hybrid systems.

https://ptolemy.eecs.berkeley.edu/~eal

Demetri Psaltis

Keynote for Session 4: Novel Physical Substrates

Professor, EPFL Lausanne, Switzerland. Some research themes: multi-mode fiber optics, optofluidics for energy conversion, nanoparticles for imaging and optofluidics, imaging in complex media, superresolution imaging. Director for the National Science Foundation Center for Neuromorphic Systems Engineering (1996–1999).

https://lo.epfl.ch/

Pieter Roelfsema

Keynote for Session 5: Guides from Neuroscience for Computing Technologies

Professor, Vrije Universiteit, Amsterdam, and Academic Medical Centre, Amsterdam; Director of the Netherlands Institute for Neuroscience, NL. Some research themes: visual perception, plasticity and memory in the visual system, behavioral paradigms in humans, computational neuroscience.

https://nin.nl/research/researchgroups/roelfsema-group/

Susan Stepney

Keynote for Session 1: Applications of Unconventional Computing Systems

Professor of Computer Science, University of York, UK. Some research themes: Non-standard computation (neural, biochemical, nanomolecular, evolutionary methods), non-von-Neumann architectures, quantum computing, artificial life, artificial immune systems.

https://www-users.cs.york.ac.uk/~susan/index.htm

Ipke Wachsmuth

Public opening lecture

Professor emeritus, former director of the Center for Interdisciplinary Research (ZiF), Bielefeld University, Germany. Some research themes: AI, cognitive learning research, natural language, expert and agent systems, virtual reality, multimodal interaction, affective computing, robotics, AI and robot ethics.

https://www.techfak.uni-bielefeld.de/~ipke/

David Wolpert

Plenary Lecture 1: Fundamental Limits on the Thermodynamics of Circuits

Member of the resident faculty at the Santa Fe Institute, USA. Some research themes: nonequilibrium statistical physics to analyze the thermodynamics of computing systems; game theory for modeling humans operating in complex engineered systems; exploiting machine learning to improve optimization; and Monte Carlo methods.

https://www.santafe.edu/people/profile/david-wolpert

Session Topics

AC - Cognitive applications of current systems

Panel Chair: Geoffrey Burr, IBM, Almaden, USA

Over the past years, technical solutions to control systems or to data analysis became increasingly cognitive. These cognitive systems unite machine learning and artificial intelligence with human interaction or brain-like information processing. This allows humans to utilize complex machine learning approaches, to make better and more informed decisions. Furthermore, brain-computer interfaces enriching or substituting human senses and behavioral control have been realized. Neuro inspired machine learning tools such as deep learning revolutionized the learning of complex data structures and act as glue within this fast growing field. Cognitive applications are the driving force behind novel machine learning concepts and would significantly profit from fully neuromorphic computation. Suggestive topics:

  • Optical devices for communication systems
  • Implantable neurochips, neuroprosthetics
  • Ultra-low power devices, embedded systems
  • Artificial agents (avatars, game characters, robots): lifelong learning and evolution
  • Distributed control systems for compliant or soft motor systems

TM - Theoretical concepts and mathematical foundations

Panel Chair: Raoul-Martin Memmesheimer, University of Bonn, Germany

An eminent obstacle opposing a fruitful transfer from neuroscience to computer science is a lack of unifying mathematical modeling tools. Different aspects of neural information processing are captured with incommensurable kinds of formal models. These tools are furthermore often highly specialized and understood by only a limited in-group of dedicated experts. Altogether novel mathematical frameworks may be needed which can capture self-organization and pattern formation in terms that connect to information processing. We solicit contributions from mathematical modelers on themes as the following indicative examples:

  • Non-autonomous (stochastic, input-driven) dynamical systems
  • Spatial / temporal multiscale modeling; nonlinear phenomena in ultra-high-dimensional inhomogeneous systems
  • "Softening" classical logic: fuzzy, stochastic,... ; metalogical frameworks
  • Non-Shannon definitions and measures of information
  • Robust phenomena in systems with less-than-1-bit precision, drifting parameters

NH - Neuromorphic hardware

Panel Chair: Thomas Van Vaerenbergh, Hewlett Packard Enterprise Labs, Palo Alto, USA

The human brain harbors billions of neurons connected via synapses to make a large-scale network. It is self-regulating, highly connected, multi-scale and consists of nonlinear devices. Computation is founded on an intrinsically parallel architecture. These features and their distinction from classical computation have stimulated the development of cognitive computing concepts. The objective of this session is to query substrate and device models and realization strategies which aim at design of "brain-like" computational systems in a wide sense. Suggestive topics of this session include

  • Digital versus analog: communication and nonlinear mapping
  • Fully parallel neuromorphic hardware concepts
  • Three dimensional architecture implementation
  • Cascadability and scalability for deep Neural Networks and incorporation into bigger systems
  • Hardware-inherent homeostasis and self-organization

NS - Novel substrates

Panel Chair: Martin Ziegler, University of Kiel, Germany

Computation is currently limited to digital computers implemented on a two-dimensional silicon substrate, wherein Boolean switching circuits are realized with transistor gates. Novel substrates – whether optical, biochemical, nano-physical or quantum effect based – will not, as a rule, admit clean transistor-like switching of binary signals. To compensate, they offer a rich spectrum of nonlinear phenomena at extremely low levels of energy consumption and extremely high speed, full 3D-parallelism and intrinsic adaptiveness. This session aims at displaying the wealth of potentially exploitable physical effects, and at hinting to device designs, in themes like

  • Electronic analog neuron models
  • Novel substrates (dielectric, spin based, ionic, optical, etc.) for neurons and synapses
  • Accelerating neuromorphic hardware via ultra-fast devices
  • Low-noise substrates for analog neuromorphic hardware
  • Substrate physics that facilitate neuron connections and signal propagation

GT - Guides from neuroscience for computing technologies

Panel Chair: Alberto Bernacchia, Cambridge University, Cambridge, UK

Biological brains thrive on the basis of a "hardware" which any computer engineer would perceive as doomed: unclocked, stochastic, heterogeneous, delay-ridden, degrading, super low-precision, with extreme device mismatches, unprogrammable, and on top of all, sometimes in need of sleep. Yet this (non-)machinery achieves information processing feats that dwarf supercomputers – and it does so in a superbly robust manner. Future artificial computing substrates optimized for energy efficiency will share many of the seemingly excruciating characteristics of neuronal wetware. This session will consider themes like

  • Information coding and processing with spiking or other stochastic dynamics
  • Homeostatic and functional stabilization mechanisms
  • Representing conceptual/symbolic knowledge in dynamical phenomena
  • Multi-timescale dynamics; memory hierarchies
  • Large-scale multi-modular systems; 3D connectivity

Schedule

Day 1 -- Dec 18th Day 2 -- Dec 19th Day 3 -- Dec 20th
Session 1(1): AC 0900 - 1030 3(1): NH 0900 - 1030 Session 5(1): GT0900 - 1030
Coffee break 1030 - 1100 Coffee break 1030 - 1100 Coffee break 1030 - 1100
Session 1(2): AC1100 - 1200 Session 3(2): NH 1100 - 1200 Session 5(2): GT1100 - 1200
Lunch break 1200 - 1330 Lunch break 1200 - 1330 Lunch break 1200 - 1330
Session 2: TM 1330 - 1600 Session 4: NS 1330 - 1600 Plenary III 1330 - 1430
Coffee discussion 1600 - 1800 Coffee discussion 1600 - 1700 Coffee discussion 1430 - 1530
Public lecture 2000 - 2130 Plenary I1700 - 1800 Poster session II1430 - 1700
  Poster session I1700 - 1930 Poster prize1730 - 1800
  Banquet and “Societal Impact Plenary”2000  
AC - Applications of unconventional computing systems
TM - Theoretical concepts and mathematical foundations
NH - Neuromorphic hardware
NS - Novel physical substrates
GT - Guides from neuroscience for computing technologies

Poster sessions

Two poster sessions will significantly broaden the scientific scope of the event. The organizing committee will not pre-define the topical orientation of posters. We want to encourage creative and out-of-the-box thinking. Poster sessions are accommodated in a generous time frame, fostering the exchange of ideas between the different communities.

The conference will award three poster prizes. Consideration for this prize will be exclusively for graduate/PhD students. A prize committee will award one prize each dedicated to (i) creativity, (ii) feasibility, and (iii) future impact.

Call for Abstracts

We solicit the submission of 2-page extended abstracts. Abstracts will be reviewed and, if accepted, be designated as an oral presentation or as a poster contribution. Please find details concerning the topics, procedures, timelines and format in the Call for Abstracts (download below).

 

Deadline for abstract submission: August 31st
Notification of acceptance: September 30th
Final abstract for online publication: October 31st


Download CallLatex and Word Templates

Abstract Submission

A selection committee will select oral (25 + 5 minutes ) and poster presentation based on the quality of the contribution. Please submit abstract for the selection committee (two pages including figures and references, the main text limited to 700 words). Evaluation will be solely based on scientific quality and relevance to the conference’s overall mission statement.

* required

Registration

Due to the strict limitation to 200 conference participants, we must use a two-stage registration process. In a pre-registration phase, open until September 30th, we are collecting applications. By September 30th, the notifications for accepted abstracts will be sent, and we will be able to finalize the registration process. Then, in the second phase, after September 30th, all first authors of accepted abstracts will be registered with priority, followed by other pre-registered participants, who will be considered in the order of their pre-registrations. Payments of registration fees will be requested only in this second phase, and registrations will become effective after receipt of the payment.

The registration interface will remain active after September 30th in the case that the admissible number of 200 participants cannot be safely reached by that date.

Registration fees:

Students (incl. PhD): 150€

Others : 300€

Industry: 900€

* required

Organizers

Dr. Daniel Brunner

FEMTO-ST Research Institute, Besançon, France.

http://members.femto-st.fr/daniel-brunner


Prof. Gordon Pipa

Institute of Cognitive Science, Osnabrück, Germany.

https://www.ikw.uni-osnabrueck.de/en/research_groups/neuroinformatics/people/prof_dr_gordon_pipa.html


Prof. Herbert Jaeger

Jacobs University, Bremen, Germany.

https://www.jacobs-university.de/directory/hjaeger


Prof. Stuart Parkin

Max Planck Institute for Microstructure Physics, Germany

http://www.mpi-halle.mpg.de/NISE/director