SoC2014 kevin AI - Revision history

KevinXi at 03:43, 17 April 2014

2014-04-17T03:43:32Z

← Older revision		Revision as of 03:43, 17 April 2014
Line 300:		Line 300:

	https://github.com/wesnoth/wesnoth/pull/144		https://github.com/wesnoth/wesnoth/pull/144

			https://github.com/wesnoth/wesnoth/pull/152

	'''3) Communication skills'''		'''3) Communication skills'''

KevinXi: /* Issues */

2014-04-16T00:22:45Z

Issues

← Older revision		Revision as of 00:22, 16 April 2014
Line 129:		Line 129:
	**The very last hitpoint of a unit should be more valuable than first few lost. This make sense because we want injured unit retreat. I am trying multiple functions to show the rise of value for each hitpoint.		**The very last hitpoint of a unit should be more valuable than first few lost. This make sense because we want injured unit retreat. I am trying multiple functions to show the rise of value for each hitpoint.
	**“you can convert gold into units, but this takes some time - 1 turn to recruit, and 1-2 more turns to move to the frontline. So, there is 'current' gold (units on the frontline) and 'future' gold (income and units not yet on the frontline)” as crab_ mentioned.		**“you can convert gold into units, but this takes some time - 1 turn to recruit, and 1-2 more turns to move to the frontline. So, there is 'current' gold (units on the frontline) and 'future' gold (income and units not yet on the frontline)” as crab_ mentioned.
	*Experience of units should take into consider		**Experience of units should take into consider
	*Currently the intermediate value of hitpoints lost in battle simulation is store in the int unit::hitpoint_, need new double variable to store the expectation of the hitpoints lost, or just run a real simulation. Find out which way is good.		*Currently the intermediate value of hitpoints lost in battle simulation is store in the int unit::hitpoint_, need new double variable to store the expectation of the hitpoints lost, or just run a real simulation. Find out which way is good.
	*The balance of each part of state(units, villages)		*The balance of each part of state(units, villages)

KevinXi: /* Issues */

2014-04-16T00:21:51Z

Issues

← Older revision		Revision as of 00:21, 16 April 2014
Line 130:		Line 130:
	**“you can convert gold into units, but this takes some time - 1 turn to recruit, and 1-2 more turns to move to the frontline. So, there is 'current' gold (units on the frontline) and 'future' gold (income and units not yet on the frontline)” as crab_ mentioned.		**“you can convert gold into units, but this takes some time - 1 turn to recruit, and 1-2 more turns to move to the frontline. So, there is 'current' gold (units on the frontline) and 'future' gold (income and units not yet on the frontline)” as crab_ mentioned.
	*Experience of units should take into consider		*Experience of units should take into consider
			*Currently the intermediate value of hitpoints lost in battle simulation is store in the int unit::hitpoint_, need new double variable to store the expectation of the hitpoints lost, or just run a real simulation. Find out which way is good.
	*The balance of each part of state(units, villages)		*The balance of each part of state(units, villages)
	*Current version do not have much difference with brute force search. How to show the advantage of dynamic programming, that is, the usage of intermediate results?		*Current version do not have much difference with brute force search. How to show the advantage of dynamic programming, that is, the usage of intermediate results?

KevinXi at 11:25, 15 April 2014

2014-04-15T11:25:12Z

← Older revision		Revision as of 11:25, 15 April 2014
Line 121:		Line 121:
	===Example Output===		===Example Output===

	http://pastebin.com/~~7rtUhM5Z~~		http://pastebin.com/wwzcHE3n

	===Issues===		===Issues===

KevinXi: /* Questionnaire */

2014-04-15T09:31:12Z

Questionnaire

← Older revision		Revision as of 09:31, 15 April 2014
Line 280:		Line 280:
	'''2.5.2) What type of games?'''		'''2.5.2) What type of games?'''

	Role-playing, Alternate reality, Educational, Puzzles, Strategy.		Role-playing(Roguelike, nethack is my favourite!), Alternate reality, Educational, Puzzles, Strategy.

	'''2.5.3) What type of opponents do you prefer?'''		'''2.5.3) What type of opponents do you prefer?'''

KevinXi: /* Questionnaire */

2014-04-15T09:27:25Z

Questionnaire

← Older revision		Revision as of 09:27, 15 April 2014
Line 347:		Line 347:
	'''4.4) Include an estimated timeline for your work on the project. Don't forget to mention special things like "I booked holidays between A and B" and "I got an exam at ABC and won't be doing much then".'''		'''4.4) Include an estimated timeline for your work on the project. Don't forget to mention special things like "I booked holidays between A and B" and "I got an exam at ABC and won't be doing much then".'''

	See [[SoC2014_kevin_AI#Plan\|"Plan"]]. I may need one week off for final exam, it usually happend at the end of June, and another ~~few days~~ to return home from university and spend time with my parents, at the beginning of summer vacation.		See [[SoC2014_kevin_AI#Plan\|"Plan"]]. I may need one week off for final exam, it usually happend at the end of June, and another week to return home from university and spend time with my parents, at the beginning of summer vacation.

	'''4.5) Include as much technical detail about your implementation as you can'''		'''4.5) Include as much technical detail about your implementation as you can'''

KevinXi at 09:25, 15 April 2014

2014-04-15T09:25:12Z

← Older revision		Revision as of 09:25, 15 April 2014
Line 347:		Line 347:
	'''4.4) Include an estimated timeline for your work on the project. Don't forget to mention special things like "I booked holidays between A and B" and "I got an exam at ABC and won't be doing much then".'''		'''4.4) Include an estimated timeline for your work on the project. Don't forget to mention special things like "I booked holidays between A and B" and "I got an exam at ABC and won't be doing much then".'''

	See [[SoC2014_kevin_AI#Plan\|"Plan"]].		See [[SoC2014_kevin_AI#Plan\|"Plan"]]. I may need one week off for final exam, it usually happend at the end of June, and another few days to return home from university and spend time with my parents, at the beginning of summer vacation.

	'''4.5) Include as much technical detail about your implementation as you can'''		'''4.5) Include as much technical detail about your implementation as you can'''
Line 388:		Line 388:

	'''5.2) Which tools do you normally use for development? Why do you use them?'''		'''5.2) Which tools do you normally use for development? Why do you use them?'''

	Vim, g++, gdb, make, git. Because these tools are build-in and powerful, and I can learn and modify them since open source. I will use IDE for some big project.		Vim, g++, gdb, make, git. Because these tools are build-in and powerful, and I can learn and modify them since open source. I will use IDE for some big project.

Line 395:		Line 396:

	'''5.4) Would you mind talking with your mentor on telephone / internet phone?'''		'''5.4) Would you mind talking with your mentor on telephone / internet phone?'''

	Not at all, I will send my phone number to my mentor if I am accepted.		Not at all, I will send my phone number to my mentor if I am accepted.

KevinXi: /* Timeline */

2014-04-15T09:18:32Z

Timeline

← Older revision		Revision as of 09:18, 15 April 2014
Line 148:		Line 148:
	! Time !! Tasks		! Time !! Tasks
	\|-		\|-
	\| 21 Apr - ~~4 May~~ \|\|		\| 21 Apr - 27 Apr \|\|
	*<b>Prepare step</b>		*<b>Prepare step</b>
	*Familiar with the overall structure of Wesnoth project		*Familiar with the overall structure of Wesnoth project
	*Familiar with the AI part		*Familiar with the AI part
	\|-		\|-
	\| ~~5 May~~ - 11 May \|\|		\| 28 Apr - 11 May \|\|
	*<b>General framework setup step</b>		*<b>General framework setup step</b>
	*Design a general framework for dynamic programming be applied to battlefield		*Design a general framework for dynamic programming be applied to battlefield

KevinXi: /* Benchmarks */

2014-04-15T09:15:58Z

Benchmarks

← Older revision		Revision as of 09:15, 15 April 2014
Line 139:		Line 139:
	*Win 70% games against default RCA AI		*Win 70% games against default RCA AI

	*~~its~~ behaviour achieving the goal of 'analyze and make decisions globally', e.g. sometimes it give up local optimal to achieve global optimal.		*Its behaviour achieving the goal of 'analyze and make decisions globally', e.g. sometimes it give up local optimal to achieve global optimal.

	==Project==		==Project==

KevinXi: /* Project */

2014-04-15T09:12:26Z

Project

← Older revision		Revision as of 09:12, 15 April 2014
Line 149:		Line 149:
	\|-		\|-
	\| 21 Apr - 4 May \|\|		\| 21 Apr - 4 May \|\|
	*~~familiar~~ with the overall structure of Wesnoth project		*<b>Prepare step</b>
	*~~familiar~~ with the AI part		*Familiar with the overall structure of Wesnoth project
			*Familiar with the AI part
	\|-		\|-
	\| 5 May - 11 May \|\|		\| 5 May - 11 May \|\|
	*~~learn dynamic programming, understand the advantages and of it~~		*<b>General framework setup step</b>
	*~~design~~ a general framework for dynamic programming be applied to battlefield		*Design a general framework for dynamic programming be applied to battlefield
	*~~write~~ design document for it		*Write design document for it
			*implement dynamic programming framework
	\|-		\|-
	\| 12 May - 18 May \|\|		\| 12 May - 25 May \|\|
	*~~implement prototype of~~ dynamic programming ~~framework~~		*<b>Dynamic programming elements setup step</b>
	\|-		*Learn more on dynamic programming, understand the advantages and of it
	~~\| 19 May - 25 May \|\|~~		*Refine the value of state(villages, units), make them comparable under one standard
	*~~find a method to test~~ the ~~result~~ of ~~new AI~~(~~maybe AI arena?~~)		*Implement the stage_state utilizing the formula figured out last week
	*~~implement it if do not have~~		*Maybe Machine Learning can help to decide the value of state
	\|-		\|-
	\| 26 May - 1 Jun \|\|		\| 26 May - 1 Jun \|\|
	*~~refine~~ the ~~value~~ of ~~state~~(~~villages, units~~)~~, make them comparable under one standard~~		*<b>Test environment setup step</b>
			*Find a method to test the result of new AI(maybe AI arena?)
			*Implement it if do not have
	\|-		\|-
	\| 2 Jun - 8 Jun \|\|		\| 2 Jun - 15 Jun \|\|
	*~~implement~~ the ~~stage_state utilizing the formula figured out last week~~		*<b>CAs setup step</b>
	\|-		*Design and write document for the design of CAs
	~~\| 9 Jun - 15 Jun \|\|~~		*Implement and test corresponding CAs, test them individually
	*~~refine~~ the value of state(villages, units)~~, make them comparable under one standard~~		*Refine the value of state(villages, units) again according to the performance of CAs
	\|-		\|-
	\| 16 Jun - 29 Jun \|\|		\| 16 Jun - 29 Jun \|\|
	*design ~~and write document for~~ the ~~design of~~ CAs		*<b>CAs "design-implement-test" iterations step</b>
	*~~implement~~ and ~~test corresponding CAs~~		*Make the CAs interact with state
			*Plug CAs into the general framework
			*Test and deal with feedback
			*Machine learning may help
	\|-		\|-
	\| 30 Jun - 6 Jul \|\|		\| 30 Jun - 6 Jul \|\|
	*make the CAs interact with state
	*plug CAs into the general framework
	\|-
	~~\| 7 Jul - 13 Jul \|\|~~
	*test AI, analyze result, balance argument		*test AI, analyze result, balance argument
	*is there a automatic way to balancing?		*is there a automatic way to balancing?
	\|-		\|-
	\| 14 Jul - 20 Jul \|\|		\| 7 Jul - 13 Jul \|\|
	*maybe public to players and collect data and respond		*maybe public to players and collect data and respond
	\|-		\|-
	\| 21 Jul - 27 Jul \|\|		\| 14 Jul - 27 Jul \|\|
	*judge if it 'global' enough		*judge if it 'global' enough
	\|-		\|-