Reinforcement Learning and Artificial Intelligence (RLAI)

Tea-time Talks 2007


Similar to last year, we will have the a tea-time talk every Monday to Thursday this summer at 4:30-5:00pm in CSC 333. Refreshments will be provided, starting at 4:15pm. 

The intention of the tea-time talks is to efficiently transmit information on a variety of current Reinforcement Learning topics.

The ambition of this page is to organise the tea-time talks and provide a mailing list for all participants.

last year's tea-time page


Guidelines

Organisation



Schedule: (Staring June 11)

Date
Presenter
Topic
Link
June 11-14
[Organiser: ]
Adam White

Mon
Csaba
NIPS submission

Tue
Mohammad
Analyzing Feature Generation for Value-Function Approximation
PDF
Wed
Cancelled
AICML Site visit

Thu
Rich
on the role of tracking in non-stationary environments
pdf
June 18 - June 22
[Organiser: andrew]
 

Mon
Alborz
NIPS submission
pdf
Tue
vgrover
 Cancelled Apprenticeship Learning via Inverse RL
pdf
Wed
martha.lednicky
 Cancelled To transfer or not to transfer
pdf
Thur
minh
 Cancelled

June 25 - June 28
[Organiser: andrew]
 

Mon
vgrover
Apprenticeship Learning via Inverse RL
pdf
Tue
martha.lednicky
To transfer or not to transfer
pdf
Wed
yakiengel "Information Value Theory" by Ronald Howard, 1966
ps
Thur
minh

July 2 - July 6
[Organiser: amir]
 

Mon
Cancelled
Holiday

Tue
mlee
 Automatic basis function construction for approximate DP and RL PDF
Wed
neufeld
 

Thur
masoud
 Disturbance RL

July 9 - July 13
[Organiser: Leah]
 

Mon
awhite
 Reinforcement Learning with Human Teachers: Evidence of Feedback and Guidance with Implications for Learning Performance
PDF
Tue
everbeek
 

Wed
elliot
 

Thur
abbasiya
 

July 16 - July 20
[Organiser: abbasiya]
 

Mon
Alborz
Linear Priotrized Sweeping

Tue
Amir massoud
Regularization, L_1 Penalty, and Forward Stagewise Linear Regression

Wed
anna
 

Thur
brian
 

July 23 - 27
[Organiser: everbeek]


Mon
arash Collaborative Multiagent Reinforcement Learning
pdf
Tue
cancelled


Wed
cancelled

Thur
Brad Joyce
Intrinsic Motivation

Jul 30 - Aug 2
[Organizer - Adam]

Mon
Leah Hackman


Tue
Sandra Zilles
Basic Concepts of Algorithmic Learning PDF
Wed
Masoud Shahamiri
Automatic shaping and decomposition of reward functions
PDF
Thur
James Neufeld
Robotic Search *Discussion*

Aug 6 - 9 [Organizer - Vlad]


Mon
Civic Holiday


Tue
canceled



Wed
Volodymyr Mnih Potential-based shaping and Q-value initialization are equivalent
PDF

Thur
Yasin Abbasi-Yadkori Variance of Response in Bayesian Networks


Aug 13 - Aug 16
[Organizer - Brad]


Mon
Cancelled


Tue
Brad Joyce

Wed
Amir massoud Farahmand
The Linear Programming Approach to Approximate Dynamic Programming (de Farias and Van Roy)

Thur
Rich Sutton


Aug 20 - Aug 23 [Organizer - Varun]


Mon
Hamid Reza Maei
Wake-Sleep Algorithm for Representational Learning (with application to real data)

Tue
Brian Tanner
Towards Parameter Free RL : A peak at Brian's thesis ideas

Aug 27 - Aug 30
[Organizer - Mark]



Mon
Marc Bellamare



Tue
Elliot Ludvig



Wed
David Silver



Thur
Adam White



Sep 3 - Sep 6
[Organizer - Vlad]



Mon
Labour Day



Tue
Varun Grover






Paper suggestions:
For conferences and JMLR, we may specify some papers explicitly. There are several relevant papers there.


Extend this page to add a suggestion, or edit to remove as appropriate. (Click on "Extend this Page" in footer).



brian extend  

Today's Tee Time talk will be in Room CS 333. The talk will start at 4:30. Please come 5 - 10 minutes early to get settled and enjoy cookies and tea. Mohammad will be presenting "Analyzing Feature Generation for Value-Function Approximation" paper.  

A good paper for our TTT is ''Automatic basis function construction for approximate DP and RL'' by Keller, Mannor, and Precup.  You  can find it at

http://imls.engr.oregonstate.edu/www/htdocs/proceedings/icml2006/057_Automatic_Basis_Func.pdf

I also have another suggestion. It'd be great if someone can talk about "Locally Weighted Regression" in our TTT.  

Mohammad

Today's Tea Time talk will be in Room CS 333. The talk will start at 4:30. Please come by 4:15 to enjoy cookies and tea. Today's presentation is by Rich. Paper is titled "On the Role of Tracking in Stationary Environments".  

I have created the schedule for next 30 days. Please have a look. If some dates do not work for you swap with someone else.

Thanks,

Varun  

Today's Tea Time talk will be in Room CS 333. The talk will start at 4:30. Please come by 4:15 to enjoy cookies and tea. Today's presentation is by Alborz.

Topic: Online Control With Least-Squares Methods (NIPS 2007 submission)

Abstract:
 Policy evaluation using least-squares techniques (such as LSTD and iLSTD) have been shown to estimate the value of a policy with far less data than traditional TD techniques.  Unfortunately, they make use of policy-dependent statistics that have to be discarded when the policy changes.  This makes it difficult to use the techniques for online control problems.  In this paper, we explore the effect of policy on the least-squares statistics, distinguishing three fundamental effects.  We then introduce the framework of least-squares Sarsa (LSS and iLSS) and empirically evaluate previously suggested approaches for handling data from older policies in the least-squares statistics.  We show these approaches can maintain the least-squares data efficiency in some control problems, identify circumstances where least-squares approaches can be problematic and where special handling of data from older policies improves learning.  

What do people think about not having TTT this week due to ICML. Some people have told me that they would rather not present during this week. Please send me your opinion. If enough people oppose the talks this week then we can move them next week.  

We will not have TTT for rest of this week due to ICML :(
I will update the schedule and move this week's talks to next weeek.  

Today's Tea Time talk will be in Room CS 333. The talk will start at 4:30. Please come by 4:15 to enjoy cookies and tea. Today's presentation is by myself. I am presenting a paper titled "Apprenticeship Learning via Inverse Reinforcement Learning".  

Today's Tea Time talk will be in Room CS 333. The talk will start at 4:30. Please come by 4:15 to enjoy cookies and tea. Today's presentation is by Martha on "To transfer or not to transfer".  

Today's Tea Time talk will be in Room CS 333. The talk will start at 4:30. Please come by 4:15 to enjoy cookies and tea. Today's presentation is by Yaki on "Information Value Theory".  

Unfortunately, CSC 333 is booked today till 5:00. Therefore, we will have the TTT in room CSC 249 instead. Sorry for the mess up.  

Hey TTTimers,

 I am to present next Wednesday but my sister is getting married this weekend and I will be out of town starting tonight and I won't be back until Tuesday. Anyway I do not have the time to prepare a talk, is anyone able to switch me? Sorry for the short notice.

-James  

TTimers,

I am off for a month. Vlad will be taking over the responsibilities of TTT for the next month. Please send your question/queries/grievances about TTT to Vlad. Thanks Vlad for volunteering.  

Please join us for tea and cookies in CS 333 at 4:15. At 4:30 I will present "Scaling Learning Algorithms towards AI" by Bengio and LeCun.  

Please join us for tea and cookies in CS 333 at 4:15. At 4:30 Mark will present "Automatic basis function construction for approximate DP and RL" by Keller, Mannor, and Precup.  

Please join us for tea and cookies in CS 333 at 4:15. At 4:30 David will give his ICML talk on "Combining Online and Offline Learning in UCT".  

Please join us for tea and cookies in CS 333 at 4:15. At 4:30 Masoud will present his work on "Disturbance reinforcement learning".  

Please join us for tea and cookies in CS 333 at 4:15. At 4:30 Sandra will talk about "Basic Concepts of Algorithmic Learning".  

Please join us for tea and cookies in CS 333 at 4:15. At 4:30 Masoud will present "Automatic shaping and decomposition of reward functions" by Bhaskara Marthi.  

There will be no tea talk today. We will continue tomorrow when I will present "Potential-based shaping and Q-value initialization are equivalent" by Eric Wiewiora.  

Dear RLAI'ers.
I just returned from vacation, and found out that I'm scheduled to give a another TT talk on the coming monday (Aug.13), in what I guess is the 2nd round of talks.
As we're leaving at the end of the month, I'm terribly busy, and won't be able to prepare a talk until monday.
I'll appreciate it if someone who hasn't given a 2nd TT talk could replace me in that spot.
Thanks, Yaki.  

Please join us for tea and cookies in CS 333 at 4:15. At 4:30 Yasin will talk about "Variance of Response in Bayesian Networks".  

There will be no teatime talk today. Also, the schedule has been updated so please have a look to see if you are giving a talk.  

Since nobody volunteered to replace me today, I actually did prepare something for my talk today. Sorry for not writing earlier!
I'll talk about tile kernels - see you at 16:30!
Yaki.  

Today, August 14th 2007, Dan will be discussing transportation and planning for our trip to Banff for the iCore. It is *imperative* that everyone going to Banff next week attend this meeting. Same tea-time, same tea-place.  

Please join us for tea and cookies in CS 333 at 4:15. At 4:30 Amir massoud will present "The Linear Programming Approach to Approximate Dynamic Programming" by de Farias and Van Roy.  

Please join us for tea and cookies in CS 333 at 4:15. At 4:30 Rich will talk about his plans for a new book.  

Please join us for tea and cookies in CS 333 at 4:15. At 4:30 Hamid will talk about using the Wake-Sleep algorithm for representational learning.  

Please join us for tea and cookies and a special suprise (!) in CS 333 at 4:15. At 4:30 Adam will give a talk.  

There will be no teatime talk today. The next talk is on Tuesday, September 4.  

Please join us for our last teatime talk of the year at 4:30 in CS 333. At 4:45 Varun will talk about various sequence models.  

Extend this Page   How to edit   Style   Subscribe   Notify   Suggest   Help   This open web page hosted at the University of Alberta.   Terms of use  2990/29