ANNCBT Questions and Answers
Artificial Neural Networks and Computational Brain Theory Group
questions and answers
INDEX
What's new on the web pages?
What is ANNCBT?
Who runs ANNCBT?
Who comes to ANNCBT meetings?
Where are the ANNCBT meetings and talks held? Directions?
How do I become a member of ANNCBT?
I'd like to address the group, what do I need to do?
Where and when does ANNCBT meet?
What's the format of the meetings?
What should I have prepared if I am leading a discussion?
What kinds of papers are appropriate for discussion?
What kind of background should I have to register in the
official 591 seminar?
What are the differences between the credit options
offered for the official 591 seminar?
I'm having trouble posting/resubscribing to the mailing list.
Why was the mailing list closed to non - university of illinois people? Why was it re-opened to the public?
What mailinglist TOPICS are there? How do I set the TOPICS i am subscribed to?
Help - I'm receiving mail from all kinds of idiots.
Help - I'm being inundated by email from the ANNCBT group at the university of illinois.
What's with that logo?
- What's new on the web pages?
Wed. sep 01/99 - the neuroscience crosslisting has been approved.
Sun. aug 29/99 - added technical reports pointer to main page. technical reports will be available soon.
Sun. mar 07/98 - announcement posted about our special half-day workshop on Adaptive Computation at UIUC.
- What is ANNCBT?
ANNCBT was formed in january of 1996, as an informal graduate/faculty
discussion and reading group. It is not a research group, and maintains no
facilities. Anyone interested in neural networks, the brain, or distributed
AI, is welcome at both our meetings and our open talks.
We hold meetings biweekly, where papers are discussed, and host talks on
alternate weeks, where researchers at the university are invited to speak on
their current work.
Beginning Fall semester 1997, an official
accredited ANNCBT 591 seminar is being offered
as an
adjunct to the group. Enrollment requires attendance at all ANNCBT meetings,
and may require the presentation of a paper or project.
- Who runs ANNCBT?
ANNCBT is a volunteer-run group, which is currently being "managed" by
Samarth Swarup (computer science) and Kiran Lakkaraju (neuroscience).
- Who comes to ANNCBT meetings?
The group consists largely of students and faculty from computer science and neuroscience type departments - we also have people from
psychology, physiology, entomology, physics, and others. People come into the group for different reasons, and with very different
backgrounds; most of us share an interest in artificial intelligence and the brain.
Because of our diverse backgrounds, we try to stay focused on key computational issues, and keep discussions as free of technical jargon
as possible. Participating in ANNCBT means making an effort to speak in language that is accessible to "the other side." This is not
easy, and we are often less than entirely successful, but it is, perhaps, the most important principle of the group:
Giving the computer science people a chance to hear neuroscience interpretations, and vice versa.
Since many of our conversations include substantial neurobiology, they can be difficult to follow for those with minimal background in
this area. This simply requires that one focus on the key ideas and principles and sift through the ancillary details. It takes a
little work, but it's helpful for all of us, because it forces us to refine and clarify our arguments. Needless to say, the same holds
for technical discussions in AI. In ANNCBT, no one need feel as if they are in over their head - all of us are in over our heads to
some degree.
- Where are ANNCBT meetings and talks held? Can you give me directions?
ANNCBT discussion are held at the Beckman Institue, which is on University ave., between Wright and Mathews.
Still lost? Try the uiuc on-line map for locations of Beckman Institute.
- How do I become a member of ANNCBT?
There are no qualifications for becoming a "member" of the Artificial Neural Networks and Computational Brain Theory Group; anyone can
join, and all are welcome. We have people who have been in the group since its inception, who come every week, and others who attend
quite infrequently. Our mailing list is the closest thing we have to an official membership list.
- I'd like to address the group, what do I need to do?
It's rare that we turn down a request to give a talk (more typically, we target people and badger them until they acquiesce). We
encourage student presentations and presentations of works-in-progress. On occasion we host software or hardware demonstrations.
If you'd like to address the group, or know of something that might interest our members, let us
know.
- Where and when does ANNCBT meet?
We meet every week, generally on Thursdays at 4pm in room 4269, Beckman Institute (the tower room on the fourth floor). Check the
Beckman calendar to make sure!
Discussions and talks are scheduled for one hour in length, but can occasionally go on for an hour-and-a-half. Both are held on the
same day of the week, at the same hour.
- What's the format of the meetings?
In the past, we voted (as a group) on a set of papers at the beginning of the semester, mainly based on titles and abstracts, and then worked our way through the list, picking a paper for discussion and then simply showing up and discussing it.
More recently, we have adopted a policy of having a single person responsible for "leading" each biweekly discussion. That person is expected to actually read the paper, and give a brief overview of it at the beginning of the meeting. The person leading the discussion is often the person who has suggested the paper to the group.
We try to have cookies and soda available at both our discussions and at our talks, and you are quite welcome to bring your lunch/dinner to the paper discussions.
- What should I have prepared if I am leading a discussion?
Remember that ANNCBT is first and foremost a discussion group - leading a paper discussion should not consist of a full hour talk. However, it's important to realize that not everyone (anyone?) will have read the paper as carefully as you, and many may not have sufficient background to understand the more subtle aspects. You might want to prepare a brief intro, perhaps 10-20 minutes in length, discussing in broad strokes the major points. Ideally, you might stress the implications for understanding or building large-scale intelligent systems.
Some papers may require longer introductions. If the material is new to most members, complicated, and interesting, you might take the opportunity to treat the discussion as a kind of interactive tutorial. Other papers might require only the briefest of introductions.
ANNCBT is a very informal group, and no one will be be surprised (or offended) if you don't understand every detail in the paper. You will not be expected to have all the answers. The best service you can provide, as a discussion leader, is to nudge people into the water, and throw them a line if they start to go under.
If you have a paper you think the group would be interested in, and you'd like to lead a discussion on it, just suggest it at the next meeting (it's okay if you're not a regular attendee).
An overhead projector is always available during discussions and talks. Slide projectors and LCD panels can be obtained on request.
- What kinds of papers are appropriate for discussion?
The ideal paper would be accessible to someone with limited background
knowledge in any one specific field, have implications for the design of
large-scale intelligent systems, and have ties to the biological nervous
system. Ideally, it should be:
- short (~5-20 pages) and concise; the shorter the better.
- provocative; will it inspire debate and lively discussion?
- accessible (at some level) to non-experts.
- on a topic that is unresolved.
- relevant to both brain and AI people.
- concrete; a model or a proposal of a concrete hypothesis.
In other words, an ideal paper will present some specific ideas that group
members can use as a launching point for expressing our individual opinions,
theories, and insights. It need not present a wholly convincing model - better
that it be wrongheaded but provoke a vigorous debate.
Remember also that members of ANNCBT have "real" work they're supposed to be doing, so it won't go over too well if you suggest a 50 page paper full of intricate mathematical formulas.
ANNCBT strongly encourages members to submit their own papers to the group,
either as the subject of a formal discussion, or for informal commentary by
individual members. In such cases, members are invited to provide copies of
papers intended for submission for publication, or even drafts of theses. Many
members have expressed a willingness to offer substantial critiques for such
works-in-progress.
- What kind of background should I have to register in the official 591 seminar?
The following might be a set of reasonable guidelines: If you have no
background in AI or the brain, and are merely curious about such
matters, you should by all means consider yourself welcome at the
meetings, but registration in the class might not be appropriate. If you
have only minimal background in these (and related) areas, the 1 hour
enrollment might be more appropriate than the 2 hour enrollment. Although
the
group consists mainly of graduate students and faculty members, advanced undergrads are also welcome.
- What are the differences between the credit options offered for the
official 591 seminar?
First, if you register for any amount of credit, it will be expected that
you attend all weekly discussions and talks, and that you read
all of the discussion papers. The general consensus so far is that
for 1 hour, nothing else will be expected of you.
The 2 hour credit option will require either that you lead a discussion of
a paper, or
give a formal talk to the group. Which one of these options is ultimately decided upon will depend on how many students sign up for the class, how many talk slots are available, what students prefer to do, etc. An additional option might be to write a final paper instead of presenting. Of course, we would always prefer to hear a talk about student research-in-progress.
Registering for 2 hours will also obligate you to write a short commentary
on
any work-in-progress submitted by another classmate/group member. Of course, you are also welcome to submit you own papers for critique.
For students looking for more than 2 hours, the attending faculty members
have
expressed a willingness to supervise a related independent study
project. In this scenario, you would be expected to register in the
ANNCBT class for 2 hours, and register (sometime into the semester) for an
additional 2 hours of independent study with the faculty member (from
ANNCBT) of
your choice. Then, sometime towards the end of the semester, you will be expected to give a talk on your project.
If you are in doubt about how much credit to sign up for, don't panic! Sit in on the first few meetings and see how you feel about it. If you'd like to work out another arrangement for credit, talk to the faculty supervisors.
- I'm having trouble posting/resubscribing to the mailing list even though i am a member, any suggestions?
Yes. The mailing list is very picky about where you send your email from. If
your adress on the mailing list is xyzzy@sparc7.cs.uiuc.edu, then the mailing
list will complain about email sent from xyzzy@sparc5.cs.uiuc.edu. The best
solution is to make sure the address on the list is as general as possible,
i.e. xyzzy@cs.uiuc.edu.
If the list keeps denying that you are subscribed, you should email me with a request to change your mailing list address.
- Why was the mailing list closed to non - university of illinois people?
Why was it re-opened to the public?
The anncbt-l mailing list was initially open to the public, but we found that
this was problematic because of the large number of posts on subjects like 'how
do i use a nural net to make lots of money?'. Beginning 2/20/98, we are going
to re-open the mailing list to the public. To combat our past problems, we are
implementing two new policies:
- all messages posted to anncbt-l will first be forwarded to a moderator who will decide whether the message is at least tangentially related to group interests.
- messages will be classified into topic areas; in order to restrict the
volume of mail you receive, you may choose to subscribe to a select subset of topics.
- What TOPICS are there? How do I set the TOPICS i am subscribed to?
By default, subscribers are set to receive messages on ALL topics. This may
make your life unpleasent.
To select the topics you subscribe to, send a message to listserv@postoffice.cso.uiuc.edu with message text:
SET anncbt-l TOPICS: topic1 topic2 topic3 ...
Where topic1, topic2, topic3, etc. are from:
- mailinglist: (official) announcements related to the mailing list.
- meeting: (official) UofI local ANNCBT meeting announcements.
- talk: (official) UofI local ANNCBT open talk announcements.
- reminder: (official) UofI local ANNCBT reminders.
- ANNCBT: (official) UofI local ANNCBT misc. group messages.
- relatedtalk: related UofI talks that might be of interest.
- uiuc: misc. posts for or from University of Illinois people.
- Other: other posts to the group
- I'm receiving mail from all kinds of idiots! All i really want to hear
about is upcoming local University of Illinois ANNCBT talks.
By default, subscribers are set to receive messages on ALL topics.
If you are a university of illinois student or faculty interested in casually
attending our local group meetings but not in receiving other messages, you may
wish to immediately remove yourself from the Other topic, by sending the
following message to listserv@postoffice.cso.uiuc.edu:
SET anncbt-l TOPICS: mailinglist, meeting, talk
(if you don't plan on attending our meetings but want to be informed about
our open talks, you would leave out the meeting topic in the above).
- I'm being inundated by email from the ANNCBT group at the university of
illinois! All i really want to talk about is neural networks and computational
brain theory.
If you are not in the vicinity of the UofI urbana-champaign campus, you might
want to remove yourself from all but the Other topic, by sending the following
message to listserv@postoffice.cso.uiuc.edu:
SET anncbt-l TOPICS: mailinglist, Other
- What's with that logo?
The logo is meant to suggest a synergy between artificial neural
networks and the brain. The cartoon image of the brain was taken from one of Nina Paley's
"Adventures of Nina" comic strips (nina was a local urbana-champaign star). [you can see the logo
dance on the home page if you're patient]. The top five mistaken
interpretations of our logo are: 5) a bunch of grapes. 4) a water stain. 3)
cauliflower/broccoli. 2) the lower intestines. 1) an anti-abortion
symbol of a fetal pig. Congratulations to the winners.
Did we not answer your question? Let us know about it.
Last updated Dec 21, 2004.