Workshop Results

The 3-week 2011 workshop had 80 participants from 22 countries. wiki:InTheNews lists news articles referring to the workshop, including  this front page Sunday feature article in the Telluride Daily Planet.

Participants were polled on their interests in the nervous system according to Terry Sejnowski's diagram showing levels of investigation. The results were as shown below:

distribution of interests

The workshop had 4 main topic areas and a several tutorial topic areas, including an invited tutorial on asynchronous computation and communication. In addition, practical methods tutorials were taught on field programmable analog arrays, surface mount assembly, USB-based embedded microcontrollers, on-chip bias generators, and event-based sensory processing using  silicon retinas and  silicon cochleas. In the second week there was a series of seminars on computational neuroscience. The main topic area results of the workshop are described in detail in the pages linked below.

This year's workshop focused on engineering networks and systems for desired cognitive functions and the inclusion of some linguistic aspects of cognition. Notably absent from this year's workshop were themes on Brain-Computer Interfaces and Mobile Autonomous Robotics, and the past focus on circuit design was present to a much lessor extent as participants instead used already-developed hardware in their projects.

The main tools learned by participants were the Nengo Neural Engineering framework, the Leabra framework, and the Neurogrid hardware platform. As usual at Telluride, a great deal of effort was put into interfacing components -- using ROS to interface sensors in the CogRob topic, interfacing between the Nao robot and Leabra, between Nengo and Neurogrid, between Nengo and jAER, between Nengo and SpiNNaker, and between silicon cochleas and retinas to the rest of the Attention system.

This workshop also featured the first use of a powerful large display tablet in the lecture room, so that 'chalk-talk' lectures and notes from discussions could be saved and put onto the wiki.

The slides from the final presentations are available at https://neuromorphs.net/nm/wiki/FinalSlides.

Highlights of technical results


  • Results from the Guided Reinforcement Learning topic area. Highlights are the first use of real (as opposed to simulated) visual input to the Leabra vision system and the first use of Leabra on the Nao robot.

  • Results from the Asynchronous Computation and Communication Tutorial are on the tutorial wiki page. Some highlights are the excellent lectures on design available on the wiki and the tools written by Rajit Manohar made available to workshop participants (albeit still closed source, owing to prior restrictions).

Video highlights


Feedback on the workshop

After the workshop,  all participants were solicited for feedback on various aspects of the workshop using this Telluride 2011 Feedback form. Questions included what participants liked, what they thought could be improved, what topics could be introduced in the future, how the workshop could be funded, whether it should also be called a summer school, and what career stage they were at.

A  summary of these responses is available and the full results are available from a link in this summary. We are extremely grateful for the fact that there were more than 32 responses to the survey, most of them providing extensive suggestions. Responses came from people at all career stages:

Participants mentioned that they liked the project-based approach of the workshop (21 mentions), it's broad focus (20 mentions), it's informality and openness and the chance to interact with faculty (15 mentions), the workshop location (11 mentions), the chance to work with people from other fields (7 mentions), and the computational neuroscience talks in the second week (4 mentions).

Participants suggested improving the workshop by more advanced preparation of the projects including making tools available for pre-installation (10 mentions), less preconceived project plans (5), less lectures or clearer specs on mandatory lectures (6), more interaction within and between topic areas (6), more interaction between students and faculty (4), and better use of existing high-performance modules for vision and audition (3). Several participants mentioned that some topics were too per-conceived and were so complex in realization that only the originating lab experts could contribute; several however also suggested that projects could be more prepared before the workshop. It is not clear how to reconcile these suggestions. Discussions about projects could be started before the workshop starts and tools could be made available more in advance.

Suggestions for future workshops included no suggestions or "just repeat - themes were great" (7 mentions), more hardware approaches (7 mentions), and other approaches such as linguistics (2), AI (2), fMRI (1), robotics (3), BCI (2).

There was a broad spectrum of suggestions for seeking future funding as shown below; the most popular choices (multiple allowed) were NSF funding (40%), self-funding (40%), or industry funding (33%):

Most participants thought that future workshops should be called ether "Bio-inspired cogntion engineering workshop" or "Neuromorphic Cognition Engineering workshop" (50%). Only 5% voted for "Bio-inspired computing". Other suggestions included "Neuromorphic Computation and Cognition Engineering workshop" and "Neuromorphic Systems Engineering Workshop". One participant mentioned however that there is already a "Biologically Inspired Cognitive Architectures" society and conference ( http://bicasociety.org/).

There was a consensus of more than 70% not to call the workshop in addition a "summer school".

Previous year's results

The results of the 2010 workshop are available at wiki:2010/Results.