print out the classification (for a regression)
normalise the means of diff number of rounds (for the t test)
the classification changes with cushions (how?)
revise the classification
| To do | Date | By | Note |
|---|---|---|---|
| spin off from sample-fsm2 | 7 Dec | chi | |
| change "automata.rkt" to fit bargaining game | 7 Dec | ||
| add classic automata | 8 Dec | ||
| plot the number of types in each cycle | x | ||
| pass test 1 and test 2 of RD | x | ||
| fix out-rank | x | ||
| run across delta | 11 dec | ||
| write a montecarlo | hm | ||
| optimize code, now time is spent in match-up and rank | |||
| write acknowledgement | x | ||
| test 3 passed | 28 dec | ||
| output to dropbox | x | ||
| draw the new automata to check | x | ||
| run in matha | x | ||
| restructure code | 30 dec | ||
| print simulation settings as title | x | ||
| able to run across N, rounds, deltas | x | ||
| two different methods of deltas | 9 jan | ||
| set states# = 100 | x | ||
| fix a bug: in deltas, in the for loop, build-random-population in each i | x | see if it solves the flat simulations | |
| merge module c & d, from now on, run raco test for each module test | x | ||
| list the decision tree of an auto | x | ||
| return the states + the payoff in each round | 14 jan | so we can see active states of two automaton in a match (in equilibrium) | |
| walk through the workflow | x | pay attention to function vector-set! | |
| write some func to investigate automaton | v | hmm, | |
| change to no-gui to run on clusters@.. | x | ||
| try ssh key | x | ||
| back up the job description (which'd be sent to cluster) | x | next time just edit that file | |
| only export data every 10 cycle (reduce data size) | x | ||
| issue: high states -> lower periods -> maybe due to inefficient learning (not quite equilibrium). so if reduce states -> has to increase cycles. not quite because how does the inefficient periods come in stay (invade the population), there must be some mutation that mess up things | not quite | run the same combination of (cycle,states) w pd to see | |
| run across deltas, pies, states# | 17 jan | ||
| investigate the automaton, but how? | eh, plot some picture on accommodating index | ||
| the code is very messy now... (out mean each cycle, plot AI, add dir so that it can run on cluster | |||
| attempt to refactor the code so it looks like organised modules | 20 jan | ||
| maybe initially, i should set only a smaller set of states to be connected | x | 20% connected initially | |
| test to see if you change the script, it messes up the simulation on cluster, because it shouldnt, the script should be loaded for one time at the begining the of the job | x | no it doesnt, the last time it happens bc YOU forget to modify the variables | |
| what about the kind of learning speed in the probability sense that everybody has chance to learn | ignore? | ||
| resurrect automata from data & analyse population state | 23 jan | ||
| personality test (w itself, l m h) | 28 jan | ||
| personality test for mixture | x | there are cases that the test doesnt comprehend (third branch) | |
| plot the personality test | |||
| plot simulation test from data | x | see the .png | |
| see the error file for rund.pbs | x | but you deleted it, i guess it's smt when the test return '() list so list->matrix throws exception | |
| plot the bubbles in runtime | x | a night is too short, all the jobs get killed in the morning, run a shorter job | |
| investigate the 6state machines | x | get some for the paper | |
| change parameter of payoff table | 15 mar | chi | first change: H vs M -> 3,0 |
| second change: L vs L -> 5,5 | this change the max possible average payoff | ||
| changes in payoff table change the personality test structure also | think about it |
_____________________
|how nice is it |
|among its own kind? |
---------------------
|
____________________________|__________________________
| | |
________|_________ ______|_______ ______|______
100% (regarding > 50% <50%
the fair equi) -------------
-------------------- -------------- |
| | ______V_______
________V_________ ______V_______ Highs or Lows
accommodate Highs? cooperate w Fair? or complicated
------------------ -------------- --------------
| |
_______|_________ ______|________________
| | | | |
| | NO | YES
NO YES (Fair does | __V_______
(highs get | badly w it) | accommodate
<50% potential) | | | Highs?
| | __V__ ALMOST -----------
| | Tough (its tolerance |
| | ----- toward Fair |
_V_________ _V_________ increases) |
|authentic | exploit Lows? | ______V_________
|Fair | ------------ __V__ | |
---------- | Bullyish NO YES
______V_______ Tough (highs gets |
| | -------- <50% potential |
YES NO | |
(get 50% as | | __V__
highs potential) | __V__ Lows
| | Bully -----
______V_______ V -----
|Accommodator| nice
-------------- Accommodator
before running, it's necessary to change the directory for the output.
-
open
basic.rkt, the CONFIGURATION is at the beginning: -
change the directory to smt like: "/Users/linhchi.nguyen/Dropbox/fsm-bar/run1/" this directory will be append before the file name, it's better to append "" (nothing) so that the output is exported in the current directory
-
then change the variables accordingly, for example: N 100, CYCLES 100000...
-
then run the file
run-in-matha.nbin mathematica (in that file, mathematica set directory as the current directory)
in general (and inside mathematica), the command to run is
raco test -s rund basic.rkt
or
racket -tm side.rkt
with -tm and no further specification, racket will evaluate the function `main defined inside the file side.rkt, raco will run the test module inside the file, (this may be necessary when there is not only one main module inside the file, for example, run in continual probability fashion or run in discount factor fashion)
- change the directory for the output into a dropbox folder (ie smt like: "/Users/linhchi.nguyen/Dropbox/fsm-bar/run1/mean.png")
(require (file "/Users/linhchi.nguyen/Documents/fsm-bar/main.rkt"))
- change directory for output: "C:/Documents and Settings/linhchi.nguyen/My Documents/Dropbox/fsm-bar/run6/mean"
(require (file "R:/fsm-bar/main.rkt"))
This is a customised version built upon the base code of Matthias Felleisen here
The initial code of this project received a lot of critical contribution by Hoang Minh Thang. At the early stage, it is from this paper that we got inspiration for the simulation workflow.
Along the way of the development of this project, the code benefits tremendously from discussions on racket mailing list here, IRC #racket here and StackExchange.
The file "csv.rkt" has external copyright condition which can be found in its own file.