(New Participation Requirement): Blog discusion after class

Folks

Starting this week, you are required to make observations on the blog
about things
learned in the class (and readings) every week. These can be questions
you had while thinking
the material over, or skeptical comments or connections between the
week's topics and other areas.

Since we are no longer having note-taking, and have also not been
doing summaries of
papers before the class, I think there hasn't been much discussion of
things done.

In my 4-level classes, I start the discussion with "blog questions"--I
thought this will happen
automatically from your side for this class. And yet, there never seem
to be any questions.

So, I thought some creative pressure should be brought to bear.

Let us see blog posts..

regards
Rao

Semester Project Status/Sumary due next Thursday 3/6

Folks

A 2-5 page summary and status of your semester project is due in
class on next Thursday 3/6.
I realize that different students are at different stages of progress
in this. Nevertheless, the
3/6 deadline is meant to give you as well as me a clear indication of
where things stand.

In case where you still not zeroed in on a specific topic, I want
your report to give enough details
on what you have done to try to zero in.

thanks
rao

Readings for next class

Here are the readings for next class

http://rakaposhi.eas.asu.edu/real-temporal-ijcai.pdf

is the main reading. The first part of it re-states the required
concurrency property we talked about yesterday
The second part discusses the decision epoch planning at a high level.

For a more complete description of one decision epoch planner, see
Sections 1 and 2 of
http://rakaposhi.eas.asu.edu/do03a.pdf

Rao

Feedback opportunity to tell me how the class is going

Folks

Now that 6 weeks are over, I thought it would be good to poll y'all
on what is working and what is not working.
Please feel free to send your comments.

You can send comments either via anonymous web mail
http://rakaposhi.eas.asu.edu/cgi-bin/mail?rao
or by bringing a printed sheet to the class on Wednesday. Note that
the webmail records your ip address--so
you may want to send it from some ip address that is generic.

Here are some things I am interested in finding out. Other comments welcome too:

1. Are the lectures too fast/too slow/too high level/too low level etc?

2. Are you actually able to connect your readings to lectures?

3. Are you reading before or after the topics are discussed?

4. Should there be more assessment (homeworks/projects etc)?

5. How is the progress towards semester project coming (qualitatively speaking--
you will have an opportunity for detailed answers anyway)

6. Are the classes interacive enough or should they encourage more discussion?
If the latter, what can we do other than stopping for questions?

7. Overall, how does it compare to other graduate level courses you took/taking
(positively or negatively)

regards
Rao

Reading for next class (Metric Temporal Planning)

Read the paper at
http://rakaposhi.eas.asu.edu/cse574/JAIRpddl.pdf
(sections 1-5).

rao

Re: CSE 574: HW 1 Qn IV

My apologies for the mess-up in the references in Qn IV. I corrected
them now online. Here it is fyi
---------------
Qn IV

For this problem, you will use the planning graph in the last level
III.b above.

IV.a. Convert the planning graph into a CSP encoding (the problem is
small enough that you can write the entire encoding down). Show a solution
for this CSP encoding, and show how it corresponds to a plan.

IV.b. Do IV.a. but with SAT encoding of the planing graph.

IV.c.
Do an "explanatory axiom" (or backward proof based) encoding
of this problem (for the same length as the planning graph you used
in the previous parts). Mark which constraints are similar, different,
stronger etc. compared to IV.b.

-------------------

On 2/19/08, Jonathan.D.Gibbs@asu.edu <Jonathan.D.Gibbs@asu.edu> wrote:
> Okay, so here is my question again.
> In parts a and b of Question 4, we are asked to encode the planning graph from "II.b" in CSP and SAT; part c, meanwhile, asks us create a third encoding and compare its constraints to "III.b." Are these intended to reference the planning graph with mutex-propagation in Question III.b (which is derived for a domain having only 2 actions and 4 variables), or the "standard mutex graph... used by planning graph" in Question II.b (which is derived for a domain with 5 actions and 5 variables)?
>
> -Jonathan Gibbs
> jondg@cox.net
> OR
> Jonathan.D.Gibbs@asu.edu
>

Reading for Replanning--online planning--execution monitoring (next class)

Here are the readings for next class:

Required

Russell & Norvig: 12.5 & 12.6

Pell et. al. Robust planning and replanning for spacecraft..
http://rakaposhi.eas.asu.edu/cse574/pell97robust.pdf

suggested
Replanning: A new perspective
by Cushing & Kambhampati

http://rakaposhi.eas.asu.edu/replan-will.pdf

Monitoring optimality during execution
by Fritz & McIlraith

http://www.cs.toronto.edu/~fritz/publications/papers/fri-mci-icaps07.pdf

Re: Optional readings for references from today's class

Should work now.

On 2/15/08, Nan Li <Nan.Li.3@asu.edu> wrote:
> Dear Dr. Rao,
>
> I cannot open the file downloaded from the first link. Could you please
> check that? Thank you.
>
> Best,
> Nan
>
>
>
> On Thu, Feb 14, 2008 at 7:32 PM, Subbarao Kambhampati <rao@asu.edu> wrote:
>
> > Here are optional readings for references made in the class
> >
> > --> A short (4 page) paper on FF (explains the enforced hill climbing
> > and other little changes it makes on top of relaxed plan heuristic)
> >
> > http://rakaposhi.eas.asu.edu/cse574/ff-aimag.pdf
> >
> >
> > -->A (10 page) paper on Fast-Downward (that detects multi-valued
> > fluents masquerading as boolean ones, and uses relaxed planning graph
> > ideas after that detection)
> >
> >
> ftp://ftp.informatik.uni-freiburg.de/documents/papers/ki/helmert-icaps04.pdf
> >
> > --> A short (3 page) paper on Macro-FF, that selectively learns
> > macros on top of FF
> >
> http://www.cs.ualberta.ca/%7Eadib/Research/Planning/ipc4Alberta.pdf
> >
> > --> A paper on learning search control rules from failures using EBL
> >
> > http://rakaposhi.eas.asu.edu/ebl-po-aaai94.pdf
> >
> > (a longer but perhaps cleaner description is
> >

http://rakaposhi.eas.asu.edu/ebljpap.pdf
> >
> > also, for relations between EBL, dependency directed backtracking in
> > CSP (no-good learning) as well as planning, see
> >
> > http://rakaposhi.eas.asu.edu/jour-ddb.pdf )
> >
> >
> > Rao
> >
>
>

Reading for next class (over-subscription/Partial Satisfaction planning)

For next class, please read

http://rakaposhi.eas.asu.edu/psp-aaai04.pdf

Rao

Optional readings for references from today's class

Here are optional readings for references made in the class

--> A short (4 page) paper on FF (explains the enforced hill climbing
and other little changes it makes on top of relaxed plan heuristic)

http://rakaposhi.eas.asu.edu/cse574/ff-aimag.pdf

-->A (10 page) paper on Fast-Downward (that detects multi-valued
fluents masquerading as boolean ones, and uses relaxed planning graph
ideas after that detection)

ftp://ftp.informatik.uni-freiburg.de/documents/papers/ki/helmert-icaps04.pdf

--> A short (3 page) paper on Macro-FF, that selectively learns
macros on top of FF
http://www.cs.ualberta.ca/%7Eadib/Research/Planning/ipc4Alberta.pdf

--> A paper on learning search control rules from failures using EBL

http://rakaposhi.eas.asu.edu/ebl-po-aaai94.pdf

(a longer but perhaps cleaner description is

http://rakaposhi.eas.asu.edu/ebljpap.pdf

also, for relations between EBL, dependency directed backtracking in
CSP (no-good learning) as well as planning, see

http://rakaposhi.eas.asu.edu/jour-ddb.pdf )

Rao

Instructions for submission of mini-project 1

Please bring a hard-copy report on the mini-project 1 to the class.

In addition, you will also be asked to upload the pddl files of your
domains onto the class wiki.

rao

Tomorrow we will do a lecture on learning techniques for planning--Reading enclosed..

Folks

In a change of plan, I decided that it would be good to do some
discussion on the status of learning techniques--specifically those
aimed at
learning additional domain knowledge to improve planning performance.
This is something that was in the back of our minds as we kept talking
about HTNs, TLPlan rules
etc.

Although the topic of learning techniques for planning, will, I hope,
make a comeback again during the semester, this is also a good time to
do spend one lecture on it.

Accordingly, the readings for tomorrow's class are:

Sungwook Yoon's paper on learning to improve relaxed plan heuristics:

http://www.public.asu.edu/~syoon10/icaps06.pdf

You can also look at the tutorial on Learning Techniques for Planning:
http://rakaposhi.eas.asu.edu/learn-plan.html

(there is even a video of an earlier version of this tutorial at
http://videolectures.net/mlss06au_kambhampati_ltp/ )

Rao

Re: Readings for the next class

The readings are still Chap 11 and Chap 10.

We continue discussion of HTN planning (chap 11) and then do control
rules--especially
temporal control rules (chap 10)

rao

On 2/11/08, Aishwarya Sivaraman <asivaram@asu.edu> wrote:
> Dr. Rao,
>
> I was just wondering if you have put up the readings for tomorrow's class. I
> could not see any updates on the blog.
>
> Best Regards,
> Aishwarya Sivaraman
>

Homework 1 released. Due in 2 weeks.

Homework 1 is available from the home page (homeworks tab)

It covers everything done until last Tuesday

It will be due back in 2 weeks.

Please take homeworks seriously; there is a possibility that we will decide
not to have separate exams.

Also, please note that you are required to do the
work yourself and not use any old solutions that you may have been
able to download
from the web. It is okay to discuss with other students, instructor
etc, but the final work must
be done by you.

Rao

Fwd: Broken link of paper mentioned in last class

sorry. the link should work now..

http://rakaposhi.eas.asu.edu/AMali99.pdf]

The slides for the talk accompanying the paper are at

http://rakaposhi.eas.asu.edu/mali-aaai99/index.htm

Rao

On Feb 7, 2008 6:17 AM, Tuan A. Nguyen <Tuan.Anh.Nguyen@asu.edu> wrote:
> Dear Rao,
> Your paper "On the utility of plan-space (causal) encoding" can not be
> downloaded from your homepage.
> Best regards,
> A.Tuan
>

Readings for Next class: Chapters 11 and 10 in Ghallab/Nau/Traverso textbook

Folks

There will be a slight change of plan for next class. Instead of
doing partial satisfaction planning, we will do
"knowledge-based" planning--with some emphasis on hierarchical task
network planning. The readings for the
class are Chapter 11 and then Chapter 10 (in that order) from Ghallab/Nau book.

Rao

the encodings paper --working URL

Thad pointed out that the URL for the encodings paper is broken.

Here is the working URL
http://citeseer.ist.psu.edu/64263.html

(the reason the oldone is broken is that citeseer at nec is no longer working)

rao

Readings for next class (as well as additional readings for yesterday)

Assuming you have already read the readings for yesterday's class,
here are additional readings for next class

http://rakaposhi.eas.asu.edu/cse574/readings.html#encodings

The second and third papers--CSP encodings and SAT encodings--are most
relevant for next class.

The first one explains the details of how to use EBL in graphplan
(that I described at a highlevel yesterday).

rao