Transfer Learning Kickoff Meeting
			17-19 October 2005
		Sheraton Palo Alto
Oct 17	Monday
0830-0900	Check-In & Continental Breakfast
0900-1000	Welcome/Admin Remarks					LINN
	Program and Meeting Objectives					SENATOR
	Agent Remarks					CERINO
	Objective: to ensure a common understanding of DARPA's objectives for the program and the meeting.
1000-1200	Team Meetings (breakouts)
	(UCB,ISLE,NRL)					RUSSELL/LANGLEY/AHA
	break as needed
	Objective: to reinforce and clarify the responsibilities of team members. Discuss team tasks, products, transfer levels, as well as programmatics and schedule.
1200-1300	Lunch
	note: guests (e.g., Gunning, Yoon) arrive in time for the next set of presentations
	Team Leader Presentations
1300-1330	UC Berkeley Team Overview					RUSSELL
1330-1400	ISLE Team Overview					FAWCETT
1400-1430	NRL Team Overview					AHA
	Objective: provide common understanding of each team's approach and plans. Each Team Leader will present a focused overview presentation of their project including:
	* Technical gist, application domains, metrics to evaluate success.
	* Distinguish transfer learning work from existing learning work.
	* Anticipated artifacts upon successful completion of work.
1430-1445	Break
	Technical Tutorials
	Objective: provide necessary technical background to understand all aspects of the program.
1445-1515	Hierarchical Bayes					JORDAN
1515-1545	Probabilistic Relational Models					KOLLER
1545-1615	Relational Reinforcement Learning					TADEPALLI
1615-1645	Hierarchical Reinforcement Learning					DIETTERICH
1645-1715	ICARUS					LANGLEY
1715-1800	Working Dinner Buffet (In Main Confererence Room)
	Technical Tutorials (Continued)
1800-1830	Markov Logic Networks					DOMINGOS
1830-1900	HTN Learning					NAU
1900-1930	Analogical Reasoning to Support Transfer Learning					FORBUS
1930-1940	Game Theory Advisor					BRAFMAN
Oct 18	Tuesday
0800-1100	Program-Wide and Internal Team Evaluations
	Objective: ensure a common understanding of the evaluation plans.
	Evaluation Frameworks
	Explain the program-wide evaluation framework, its current status and functionality. Discuss what will be provided to each team and what will be needed from each team, along with anticipated dates.
0800-0830	TIELT					MOLINEAUX
0830-0900	GGP					LOVE
	Evaluation Applications
	Identify all the application domains for the Transfer Learning Program and how they will be used in evaluations.
0900-0915	Urban Terror					HOLDER
0915-0930	Stratagus					FERN
0930-0945	Vision					LOZANO-PEREZ
0945-1015	Break
1015-1030	Manipulation					NG
1030-1045	Physics					KYLLONEN
1045-1100	RoboCup					MACLIN/STONE
1100-1200	Transition Ideas					GREIFF
	Objective: identify potential transition avenues for TL technology.
1200-1300	Lunch
	Investigator Presentations
	Purpose: Clearly specify contributions of each Investigator's effort to the project.
	Identify approach, products, plans, metrics, milestones, schedule.
1300-1400	ISLE	(6 presentations @ 10 minutes each)
	Acquiring and Refining Relational Knowledge					MOONEY
	from Experts
	Inferring Causal Models from Background Knowledge					PAZZANI
	and Expertise
	Using Background Knowledge to Support					WITBROCK
	Transfer Learning
	Mapping Between Representations					GENESERETH
	Relational Feature Construction for
	Reinforcement Learning					LAIRD
	Reinforcement Learning for General Game Playing					STONE
1400-1500	UCB	(8 presentations @ 7 minutes each)
	Strong priors and expressive languages for					JORDAN
	hierarchical Bayes
	Transfer by generalization in decision environments					TADEPALLI
	Transfer by subtask learning in decision environments					DIETTERICH
	Theoretical foundations and methodology of					JAAKKOLA
	transfer learning
	Transfer of computational strategies by					RUSSELL
	metalevel learning
	Transfer learning in real-time strategy games					FERN
	Transfer learning in simulated robotic manipulation					KAELBLING
	Transfer learning in visual perception					KOLLER
1500-1530	Break
1530-1630	NRL	(7 presentations of varying lengths)
	1530-1542: Challenge Problem Design Process						FITZGERALD
	1542-1554: Challenge Problem Analyses						CARPENTER
	1554-1606: Learning by Doing and its Role in Testing						OATES
	1606-1612: TIELT Advice Interface						MACLIN
	1612-1618: TIELT Commercial Real-Time Strategy Game API						CONTI
	1618-1624 TIELT Commercial Turn-Based Strategy Game API						Muñoz-Avila
	1624-1630 Contribution to ISLE's project						Muñoz-Avila
	Related Projects
	Purpose: Determine how related efforts complement and contribute to projects.
1630-1650	U Wisconsin					SHAVLIK
1650-1710	Ga Tech					ISABELL/RAM
1710-1730	Rutgers					LITTMAN
1750-1800	Day 2 Wrap-Up
	Guests depart after this session
1800	Break (travel to team dinner venues if needed)
1815-1945	Team Meetings over dinner (UCB, ISLE, NRL - TBD)
	Objective: fine-tune the roles, responsibilities and schedules of each Team.
Oct 19	Wednesday
0800-0900	Evaluation Planning				AHA/COHEN/FAWCETT
	Metrics, Protocols, Schedules
	Objective: to more completely and clearly specify metrics, process, and schedule for the performance of the evaluations. Identify measures each team will use to evaluate their own TL level progress as well as the Program level (Challenge Problem) metrics.
0900-1045	Parallel Break-out Sessions for Evaluation
	Objective: Define Challenge Problem design, the testing protocol, required supporting documentation and tools that will be needed. Develop schedule. Identify specific responsibilities. Provide feedback to evaluation teams from technology developers.
	TIELT & Urban Terror					FITZGERALD
	sub breakout
		Stratagus				MOLINEAUX
		RoboCup Soccer Simulation				MACLIN OR STONE
		Vision				LOZANO-PEREZ
		Robotic Manipulations				NG
		Urban Terror				HOLDER
		Freeciv				FORBUS OR HINRICHS
	GGP					GENESERETH
	Physics					KYLLONEN
1045-1100	Break
1100-1145	Break Out Sessions Reports
	TIELT					VAN LENT
	GGP					FAWCETT
	Physics					FORBUS
	Objective: Ensure common understanding of breakout products.
1145-1200	Wrap-up/Action Items/Issues