diff --git a/Collective-Risk_Dilemma.py b/Collective-Risk_Dilemma.py
deleted file mode 100755
index e8bc02c750a20c694d5189a2842b2b1b8ac5bedd..0000000000000000000000000000000000000000
--- a/Collective-Risk_Dilemma.py
+++ /dev/null
@@ -1,335 +0,0 @@
-# NOTES:
-#
-
-
-# Imports ______________________________________________________
-
-import matplotlib.pyplot as plt
-import numpy as np
-import math
-
-
-# Constants ____________________________________________________
-
-N = 100 # Size of the population
-M = 6 # Number of individuals (from the population)
-R = 10 # Total number of rounds in one game
-T = M * R # Target Sum
-GEN = 10 # Number of generations
-G = 1000 # Number of games per generation
-PRISK = 0.9
-BETA = 1 # Used in fitness (measures the intensity of selection)
-MU = 0.03 # Error probability
-SIGMA = 0.15 # Standard deviation for Gaussian noise on thresholds
-DELTA = 0.0 # Interest
-
-# Code _________________________________________________________
-
-class Collective_Risk_Dilemma():
- def __init__(self, prisk=PRISK, sigma=SIGMA):
- self.prisk = prisk
- self.sigma = sigma
-
- self.payoffs = [[] for i in range(N)]
- self.fitness = [0] * N
- self.commonPool = 0
- self.moneyGiven = [[] for i in range(M)]
-
-
- #这个东西是用来统计 多个玩家的游戏中，一次游戏（不是一轮）结束后，系统的期待应该是
- # 每个人都能在每一轮投入1 那么应该所有人都是 cEqualsR 实际不可能， 就看看实际的情况
- # 有多少人是 free rider ，投入了但是不够， fair share, 还有无私利他的人
- self.cContributions = [0]*4 #[cEqualsZero, cEqualsR, cBiggerThanR, cSmallerThanR]
- # Set Players Strategies:
- self.contributions = np.random.choice(3, size=(N,R,2)) # 0 Defectors, 1 Fair Sharers, 2 Altruists
-
- # 结果是一个大tuple 里面有 N个小tuple 每个小tuple里面有R个元素 每个元素都是0-1之间
- self.threshold = np.random.random(size=(N,R))
- #for n in range(N):
- # self.contributions[n] = np.random.choice(3, (R, 2))
- # self.threshold[n] = np.random.random(size=R)
-
- def ComputePayoff(self, player, invested):
- """
- At the end of each game, the payoff must be recalculated
- """
- if self.commonPool >= T or np.random.random() < np.random.choice([True, False], 1, True, [1 - self.prisk, self.prisk]):
- return (2 * R) - invested
- else:
- return 0
-
- def SelectPlayers(self):
- """
- Randomly select M players from population of size N.
- """
- return np.random.choice(N, M)
-
- def UpdatePopulation(self):
- """
- The next generation is selected using the Wright-Fisher process
- where the individual’s fitness is used to weigh the probability of
- choosing an individual for the new population.
- """
- prob = list(map(lambda x: x / sum(self.fitness), self.fitness))
- reproduction_selection = np.random.choice(N, N, True, prob)
- temp_contributions = np.zeros((N,R,2))
- temp_threshold = np.zeros((N,R))
- for i, player in enumerate(reproduction_selection): # Generate errors
- for r in range(R):
- if np.random.random() < MU: # Error on contribution 1
- temp_contributions[i,r,0] = np.random.choice(3)
- else:
- temp_contributions[i,r,0] = np.copy(self.contributions[player,r,0])
- if np.random.random() < MU: # Error on contribution 2
- temp_contributions[i,r,1] = np.random.choice(3)
- else:
- temp_contributions[i,r,1] = np.copy(self.contributions[player,r,1])
- if np.random.random() < MU: # Error on threshold
- temp_threshold[i][r] = np.random.normal(self.threshold[player,r], self.sigma)
- else:
- temp_threshold[i][r] = np.copy(self.threshold[player,r])
- self.contributions = np.copy(temp_contributions)
- self.threshold = np.copy(temp_threshold)
- self.payoffs = [[] for i in range(N)]
-
- def Play(self):
- """
- Plays 1 Game (R rounds)
- """
- selectedPlayers = self.SelectPlayers() # Index of the player in the whole population (M)
- moneyPlayerOwns = [2 * R] * M # An individual player starts each game with an initial endowment of 2R
- c = [0] * M # Total Investment of each player
- self.moneyGiven = [[] for i in range(M)] # 要M个 []
- self.commonPool = 0
- for r in range(R): # 对每一轮游戏来说
- if self.commonPool < T: # If the goal was not reached yet: Put more money; Else: Do Nothing
- for i, player in enumerate(selectedPlayers):
- if self.commonPool / T < self.threshold[player,r]: # threshold not attained yet
- if self.contributions[player,r,0] <= moneyPlayerOwns[i]: # enough money to put in the commonpool
- c[i] += self.contributions[player,r,0]
- self.commonPool += self.contributions[player,r,0]
- self.moneyGiven[i].append(self.contributions[player,r,0])
- moneyPlayerOwns[i] -= self.contributions[player,r,0]
- else: # Else: Put what's left
- c[i] += moneyPlayerOwns[i]
- self.moneyGiven[i].append(moneyPlayerOwns[i])
- self.commonPool += moneyPlayerOwns[i]
- moneyPlayerOwns[i] = 0
-
- else: # threshold was attained
- if self.contributions[player,r,1] <= moneyPlayerOwns[i]: # enough money to put in the commonpool
- c[i] += self.contributions[player,r,1]
- self.commonPool += self.contributions[player,r,1]
- self.moneyGiven[i].append(self.contributions[player,r,1])
- moneyPlayerOwns[i] -= self.contributions[player,r,1]
- else: # Else: Put what's left
- c[i] += moneyPlayerOwns[i]
- self.moneyGiven[i].append(moneyPlayerOwns[i])
- self.commonPool += moneyPlayerOwns[i]
- moneyPlayerOwns[i] = 0
- self.commonPool += self.commonPool*DELTA
-
- for i, player in enumerate(selectedPlayers):
- self.payoffs[player].append(self.ComputePayoff(i, c[i]))
-
- # cContributions = [cEqualsZero, cEqualsR, cBiggerThanR, cSmallerThanR]
- for i in range(M):
- if c[i] == 0:
- self.cContributions[0] += 1
- elif c[i] == R:
- self.cContributions[1] += 1
- elif c[i] > R:
- self.cContributions[2] += 1
- elif c[i] < R:
- self.cContributions[3] += 1
- else: raise Exception
-
-
- if T < self.commonPool: return 1 # If target was reached
- else: return 0
-
-
-
-
-
-
-
- #def GetPayoffs(self):
- # return self.payoffs
-
-
- def GetAverageFitness(self):
- return np.mean(self.fitness)
-
-
- def GetAveragePayoff(self):
- payoff_sum = 0
- for n in range(N):
- payoff_sum += np.mean(self.payoffs[n])
- return np.mean(payoff_sum)
-
- def GetCommonPool(self):
- return self.commonPool
-
- def GetcContributions(self):
- return self.cContributions
-
-
-
- def UpdateFitness(self):
- """
- Receives a list of all the payoffs
- of a player. Computes the average
- and returns the exp(Bpii)
- """
- for n in range(N):
- self.fitness[n] = math.exp(BETA * np.mean(self.payoffs[n]))
-
-
-
-
-# Plots ________________________________________________________
-
-
-def SampleTrajectories():
- # Variables for the plots:
- ratioAveragePayoff = []
- ratioTargetReached = []
- ratioAverageContribution = []
-
- game = Collective_Risk_Dilemma() # Initialize one simulation
- for g in range(GEN):
- print("Playing Generation", g)
- timesReachedTarget = 0
- for i in range(G):
- timesReachedTarget += game.Play() # Play one game (10 rounds)
-
- ratioAveragePayoff.append(game.GetAveragePayoff() / (2*G)) # 2*G is the max payoff (to obtain a value between 0 and 1)
- ratioTargetReached.append(timesReachedTarget / G)
- ratioAverageContribution.append(game.GetCommonPool() / (2*N))
-
- game.UpdateFitness()
- game.UpdatePopulation() # Create a new population based on fitness
-
- plt.plot([x for x in range(len(ratioAveragePayoff))], ratioAveragePayoff, label="Payoff")
- plt.plot([x for x in range(len(ratioTargetReached))], ratioTargetReached, label="Target Reached")
- plt.plot([x for x in range(len(ratioAverageContribution))], ratioAverageContribution, label="Contribution")
- plt.xlabel('Generation')
- plt.ylabel('Proportion')
- plt.title('Sample trajectories for the evolutionary dynamics in collective-risk dilemmas')
- plt.legend()
- plt.show()
-
-
-
-
-
-
-
-
-def SummaryEvolutionaryDynamics1():
-
- risks = [i*(1/20) for i in range(21)]
- payoffRisk = []
- contributionRisk = []
- targetRisk = []
- firstHalf = []
- secondHalf = []
-
- for k in risks:
- print("PRISK", k)
- ratioPayoff = []
- ratioTarget = []
- ratioContribution = []
- game = Collective_Risk_Dilemma(k) # (PRISK, SIGMA)
- for g in range(GEN):
- print("Playing Generation", g)
- timesReachedTarget = 0
- for i in range(G):
- timesReachedTarget += game.Play() # Play one game (10 rounds)
-
- ratioPayoff.append(game.GetAveragePayoff() / (2*G))
- ratioTarget.append(timesReachedTarget / G)
- ratioContribution.append(game.GetCommonPool() / (2*N))
-
- game.UpdateFitness()
- game.UpdatePopulation() # Create a new population based on fitness
-
- payoffRisk.append(np.mean(ratioPayoff))
- contributionRisk.append(np.mean(ratioContribution))
- targetRisk.append(np.mean(ratioTarget))
- firstHalf.append(np.mean(ratioContribution[:len(ratioContribution)//2]))
- secondHalf.append(np.mean(ratioContribution[len(ratioContribution)//2:]))
-
- plt.plot(risks, payoffRisk, label="Payoff")
- plt.plot(risks, contributionRisk, label="Contribution")
- plt.plot(risks, targetRisk, label="Target")
- plt.plot(risks, firstHalf, label="1st Half")
- plt.plot(risks, secondHalf, label="2nd Half")
- plt.xlabel('Risk Probability')
- plt.ylabel('Proportion')
- plt.title('Summary of the evolutionary dynamics in collective-risk dilemmas')
- plt.legend()
- plt.show()
-
-
-
-
-
-
-def SummaryEvolutionaryDynamics2():
-
- risks = [i*(1/20) for i in range(21)]
- cEqualsZero, cEqualsR, cBiggerThanR, cSmallerThanR = [],[],[],[]
-
- for k in risks:
- print("PRISK", k)
- game = Collective_Risk_Dilemma(k) # (PRISK, SIGMA)
- for g in range(GEN):
- print("Playing Generation", g)
- timesReachedTarget = 0
- for i in range(G):
- timesReachedTarget += game.Play() # Play one game (10 rounds)
-
- game.UpdateFitness()
- game.UpdatePopulation() # Create a new population based on fitness
-
- # cContributions = [cEqualsZero, cEqualsR, cBiggerThanR, cSmallerThanR]
- contribution = game.GetcContributions()
- totalContributions = 0
- for cont in contribution:
- totalContributions += cont
-
- cEqualsZero.append(contribution[0]/totalContributions)
- cEqualsR.append(contribution[1]/totalContributions)
- cBiggerThanR.append(contribution[2]/totalContributions)
- cSmallerThanR.append(contribution[3]/totalContributions)
-
- plt.plot(risks, cEqualsZero, label="C = 0")
- plt.plot(risks, cEqualsR, label="C = R")
- plt.plot(risks, cBiggerThanR, label="C > R")
- plt.plot(risks, cSmallerThanR, label="C < R")
- plt.xlabel('Risk Probability')
- plt.ylabel('Frequency of the Behaviour')
- plt.title('Summary of the evolutionary dynamics in collective-risk dilemmas')
- plt.legend()
- plt.show()
-
-
-
-
-
-
-# ______________________________________________________________
-
-def main():
- #SampleTrajectories()
- # SummaryEvolutionaryDynamics1()
- SummaryEvolutionaryDynamics2()
-
-
-
-if __name__ == '__main__':
- main()
-
-
diff --git a/Stability_of_behaviors.py b/Stability_of_behaviors.py
deleted file mode 100755
index 0ee02ca4b553f6269eea902a8728402b8da38e27..0000000000000000000000000000000000000000
--- a/Stability_of_behaviors.py
+++ /dev/null
@@ -1,230 +0,0 @@
-# NOTES:
-#
-
-
-# Imports ______________________________________________________
-
-import matplotlib.pyplot as plt
-import numpy as np
-import math
-
-# Constants ____________________________________________________
-
-N = 100 # Size of the population
-M = 6 # Number of individuals (from the population)
-R = 10 # Total number of rounds in one game
-T = M * R # Target Sum
-GEN = 200 # Number of generations
-G = 1000 # Number of games per generation
-PRISK = 0.9
-BETA = 1 # Used in fitness (measures the intensity of selection)
-MU = 0.03 # Error probability
-SIGMA = 0.15 # Standard deviation for Gaussian noise on thresholds
-
-# Code _________________________________________________________
-
-class Collective_Risk_Dilemma():
- def __init__(self, robustness=False, prisk=PRISK, sigma=SIGMA):
- self.prisk = prisk
- self.sigma = sigma
-
- self.payoffs = [[] for i in range(N)]
- self.fitness = [0] * N
- self.commonPool = 0
- self.moneyGiven = [[] for i in range(M)]
- self.cContributions = [0]*4 #[cEqualsZero, cEqualsR, cBiggerThanR, cSmallerThanR]
- # Set Players Strategies:
-
- self.contributions = []
- for i in range(N):
- self.contributions.append([])
- choice = np.random.choice(3)
- for k in range(R//2):
- self.contributions[i].append(choice)
- choice = np.random.choice(3)
- for l in range(R-R//2):
- self.contributions[i].append(choice)
-
-
- def ComputePayoff(self, player, invested):
- """
- At the end of each game, the payoff must be recalculated
- """
- if self.commonPool >= T or np.random.random() < np.random.choice([True, False], 1, True, [1 - self.prisk, self.prisk]):
- return (2 * R) - invested
- else:
- return 0
-
- def SelectPlayers(self):
- """
- Randomly select M players from population of size N.
- """
- return np.random.choice(N, M)
-
- def UpdatePopulation(self):
- """
- The next generation is selected using the Wright-Fisher process
- where the individual’s fitness is used to weigh the probability of
- choosing an individual for the new population.
- """
- prob = list(map(lambda x: x / sum(self.fitness), self.fitness))
- reproduction_selection = np.random.choice(N, N, True, prob)
- temp_contributions = np.zeros((N,R,1))
- temp_threshold = np.zeros((N,R))
- for i, player in enumerate(reproduction_selection): # Generate errors
- for r in range(R):
- if np.random.random() < MU: # Error on contribution 1
- temp_contributions[i][r] = np.random.choice(3)
- else:
- temp_contributions[i][r] = np.copy(self.contributions[player][r])
- self.contributions = np.copy(temp_contributions)
- self.payoffs = [[] for i in range(N)]
-
- def Play(self):
- """
- Plays 1 Game (R rounds)
- """
- selectedPlayers = self.SelectPlayers() # Index of the player in the whole population (M)
- moneyPlayerOwns = [2 * R] * M # An individual player starts each game with an initial endowment of 2R
- c = [0] * M # Total Investment of each player
- self.moneyGiven = [[] for i in range(M)]
- self.commonPool = 0
- for r in range(R):
- if self.commonPool < T: # If the goal was not reached yet: Put more money; Else: Do Nothing
- for i, player in enumerate(selectedPlayers):
- if self.contributions[player][r] <= moneyPlayerOwns[i]: # enough money to put in the commonpool
- c[i] += self.contributions[player][r]
- self.commonPool += self.contributions[player][r]
- self.moneyGiven[i].append(self.contributions[player][r])
- moneyPlayerOwns[i] -= self.contributions[player][r]
- else: # Else: Put what's left
- c[i] += moneyPlayerOwns[i]
- self.moneyGiven[i].append(moneyPlayerOwns[i])
- self.commonPool += moneyPlayerOwns[i]
- moneyPlayerOwns[i] = 0
-
-
- for i, player in enumerate(selectedPlayers):
- self.payoffs[player].append(self.ComputePayoff(i, c[i]))
-
- # cContributions = [cEqualsZero, cEqualsR, cBiggerThanR, cSmallerThanR]
- for i in range(M):
- if c[i] == 0:
- self.cContributions[0] += 1
- elif c[i] == R:
- self.cContributions[1] += 1
- elif c[i] > R:
- self.cContributions[2] += 1
- elif c[i] < R:
- self.cContributions[3] += 1
- else: raise Exception
-
-
- if T < self.commonPool: return 1 # If target was reached
- else: return 0
-
-
- def UpdateFitness(self):
- """
- Receives a list of all the payoffs
- of a player. Computes the average
- and returns the exp(Bpii)
- """
- for n in range(N):
- self.fitness[n] = math.exp(BETA * np.mean(self.payoffs[n]))
-
-
- def StillStrategy00(self):
- return [0]*R in self.contributions
-
- def StillStrategy11(self):
- return [1]*R in self.contributions
-
- def StillStrategy22(self):
- return [2]*R in self.contributions
-
- def StillStrategy02(self):
- return [0]*(R//2)+[2]*(R//2) in self.contributions
-
- def StillStrategy20(self):
- return [2]*(R//2)+[0]*(R//2) in self.contributions
-
-
-
-
-
-
-# Plots ________________________________________________________
-
-
-def plot():
- risks = [i*(1/20) for i in range(21)]
-
- cntl00, cntl11, cntl22, cntl02, cntl20 = [],[],[],[],[]
-
-
- for k in risks:
- flag00, flag11, flag22, flag02, flag20 = True, True, True, True, True
- cnt00, cnt11, cnt22, cnt02, cnt20 = 0, 0, 0, 0, 0
- print("PRISK", k)
- game = Collective_Risk_Dilemma(k) # (PRISK, SIGMA)
- for g in range(GEN):
- print("Playing Generation", g)
- timesReachedTarget = 0
- for i in range(G):
- timesReachedTarget += game.Play() # Play one game (10 rounds)
-
- if flag00 and game.StillStrategy00(): cnt00 += 1
- else: flag00 = False
-
- if flag11 and game.StillStrategy11(): cnt11 += 1
- else: flag11 = False
-
- if flag22 and game.StillStrategy22(): cnt22 += 1
- else: flag22 = False
-
- if flag02 and game.StillStrategy02(): cnt02 += 1
- else: flag02 = False
-
- if flag20 and game.StillStrategy20(): cnt20 += 1
- else: flag20 = False
-
- game.UpdateFitness()
- game.UpdatePopulation() # Create a new population based on fitness
-
- cntl00.append(cnt00)
- cntl11.append(cnt11)
- cntl22.append(cnt22)
- cntl02.append(cnt02)
- cntl20.append(cnt20)
-
-
- plt.plot(risks, cntl00, label="C = 0000000000")
- plt.plot(risks, cntl11, label="C = 1111111111")
- plt.plot(risks, cntl22, label="C = 2222222222")
- plt.plot(risks, cntl02, label="C = 0000022222")
- plt.plot(risks, cntl20, label="C = 2222200000")
-
-
- plt.xlabel('Risk Probability')
- plt.ylabel('Generations')
- plt.title('Stability of behaviors in a collective-risk dilemma')
- plt.legend()
- plt.show()
-
-
-
-
-
-
-# ______________________________________________________________
-
-def main():
- plot()
-
-
-
-if __name__ == '__main__':
- main()
-
-
diff --git a/game.py b/game.py
index 7a93bf88a72384ded6f99b8718d9940b6d050ad2..f01410c6bf2b8c15dad04b8b0fa69b9a55f78451 100644
--- a/game.py
+++ b/game.py
@@ -29,7 +29,7 @@ class Game:
self.iterations = I
"""
- | 2-Player Game Graph Model:
+ | 2-Player Game Graph Model: (Small-world network)
|
| P: Probability of rewiring each original edge in the graph
|
@@ -115,9 +115,11 @@ class Game:
for p in [i, j]:
if np.random.uniform(0, 1) < risk:
- lossTable[p, r] += self.alpha / self.graph.getNodesNumber()[p]
+ lossTable[p, r] += self.alpha / \
+ self.graph.getNodesNumber()[p]
for i in range(self.N):
- self.players[i].updateReward(r, actionTable[i][r], lossTable[i][r])
+ self.players[i].updateReward(r, actionTable[i][r],
+ lossTable[i][r])
for r in range(self.R):
unique, count = np.unique(strategyTable[r], return_counts=True)
diff --git a/main.py b/main.py
index b2db86b1688ce22842d87c76d4bf6179be223acd..923934ef2b8330037d494ab27f103b8272a8c9fe 100644
--- a/main.py
+++ b/main.py
@@ -1,6 +1,6 @@
"""
This file contains graph methods and t-test implementations. The main
-function should produce all Figures and t-test results as the thesis.
+function should produce all Figures and t-test results in the thesis.
Author: Liyao Zhu liyao@student.unimelb.edu.au
Date: Apr. 2019
@@ -340,112 +340,31 @@ def main():
# titleComment="0.999 - decrease", legendLoc='lower left')
+ """T-tests 1. Average contribution of different T"""
+ # t_test(30, Actions, alpha=1, RF=2, threshold=(0.9, 1.0)) #p=0.2708
+ # t_test(30, Actions, alpha=1, RF=2, threshold=(0.8, 1.0)) #p=0.1096
+ # t_test(30, Actions, alpha=1, RF=2, threshold=(0.7, 1.0)) #p=0.1633
+ # t_test(30, Actions, alpha=1, RF=2, threshold=(0.6, 1.0)) #p=0.2208
+ # t_test(30, Actions, alpha=1, RF=2, threshold=(0.5, 1.0)) #p=2.2067e-08
- """
- Graph1: Number of Actions of Round r (start by 0) by Iteration
- """
-
- # RepeatTimes = 30
- # for N in [5, 10, 20, 50, 100]:
- # K = N - 1
- # for R in [1, 2, 4]:
- # for alpha in [0.2, 0.4, 0.6, 0.8, 1]:
- # data = rep(RepeatTimes, R, Actions, I, N=N, K=K, alpha=alpha)
- # for r in range(R):
- # stackPlot(data, r, Actions, I, titleComment="N="+ str(N) + ", R=" + str(R) + ", alpha=" +str(alpha) + ", Well-Mixed graph")
-
-
- # for k in [2, 99]:
- # for p in [0.8]:
- # data = rep(repeat=30, N=100, K=k, Actions=Actions, R=1, I=I, P=p)
- # stackPlot(data, r=0, Iterations=I, Actions=Actions, titleComment=("K=" + str(k) + ", P=" + str(p)))
-
- """
- Graph2: Average contribution by K, P
- """
- # graph_kp3d(Actions)
-
- """
- Graph3: Comparing a parameter (put in a list)
- """
- # stackBar(0, Actions, repeat=1, N=[5, 10, 20, 50, 100], threshold=0.6, RF=2)
- # stackBar(0, Actions, repeat=30, RF=2, threshold=[0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1])
-
- """
- Graph4: Actions by different epsilon with multi-arm bandit algorithms
- """
-
-
- # stackBar(0, Actions, repeat=30, multiArm='greedy', legendLoc='lower right',
- # epsilon=[0.05, 0.1, 0.15, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9],
- # threshold=0.3, RF=2)
-
-
- # data = rep(repeat=30, Actions=Actions, multiArm='greedy', epsilon=0.05)
- # stackPlot(data, r=0, Iterations=I, Actions=Actions, titleComment="0.05 - greedy",legendLoc='lower left')
- #
- # data = rep(repeat=30, Actions=Actions, multiArm='greedy', epsilon=0.1)
- # stackPlot(data, r=0, Iterations=I, Actions=Actions, titleComment="0.1 - greedy",legendLoc='lower left')
- #
- # data = rep(repeat=30, Actions=Actions, multiArm='greedy', epsilon=0.2)
- # stackPlot(data, r=0, Iterations=I, Actions=Actions, titleComment="0.2 - greedy",legendLoc='lower left')
-
-
-
- # data = rep(repeat=30, Actions=Actions, multiArm='greedy', epsilon=0.6, I=5000)
- # stackPlot(data, r=0, Iterations=5000, Actions=Actions, titleComment="0.6 - greedy", legendLoc='lower left')
-
- # data = rep(repeat=30, Actions=Actions, multiArm='decrease', epsilon=0.9)
- # stackPlot(data, r=0, Iterations=I, Actions=Actions, titleComment="0.9 - decrease",legendLoc='lower left')
-
- # data = rep(repeat=30, Actions=Actions, multiArm='decrease', epsilon=0.99)
- # stackPlot(data, r=0, Iterations=I, Actions=Actions, titleComment="0.99 - decrease",legendLoc='lower left')
- #
- # data = rep(repeat=30, Actions=Actions, multiArm='decrease', epsilon=0.999)
- # stackPlot(data, r=0, Iterations=I, Actions=Actions, titleComment="0.999 - decrease",legendLoc='lower left')
-
-
-
-
-
-
-
-
- """
- T-Test
- """
-
- # t_test(30, Actions, alpha=1, RF=2, threshold=(0.2, 0.3), byThreshold=True) #p=3.324e-31
- # t_test(30, Actions, alpha=1, RF=2, threshold=(0.6, 1.0)) #pvalue=0.2208
- # t_test(30, Actions, alpha=1, RF=2, threshold=(0.8, 1.0)) #pvalue=0.1096
- # t_test(30, Actions, alpha=1, RF=2, threshold=(0.5, 1.0)) #pvalue=2.2067e-08
- # t_test(30, Actions, alpha=0.85, RF=2, threshold=(0.2, 0.3), byThreshold=True) #pvalue=0.005865
-
- # t_test(30, Actions, alpha=(1, 0.9), RF=2, threshold=0.2) #pvalue=0.3748
- # t_test(30, Actions, alpha=(1, 0.85), RF=2, threshold=0.2) #pvalue=0.001466
- # t_test(30, Actions, alpha=(1, 0.8), RF=2, threshold=0.2) #pvalue=0.0002030
- # t_test(30, Actions, alpha=(1, 0.75), RF=2, threshold=0.2) #pvalue=3.9617e-07
- # t_test(30, Actions, alpha=(1, 0.7), RF=2, threshold=0.2) #pvalue=2.2428e-09
- # t_test(30, Actions, alpha=(1, 0.65), RF=2, threshold=0.2) #pvalue=6.8966e-09
- # t_test(30, Actions, alpha=(1, 0.6), RF=2, threshold=0.2) #pvalue=7.1621e-15
- # t_test(30, Actions, alpha=(1, 0.5), RF=2, threshold=0.2) #pvalue=4.1760e-13
- # t_test(30, Actions, alpha=(1, 0.45), RF=2, threshold=0.2) #pvalue=1.3749e-11
- # t_test(30, Actions, alpha=(1, 0.4), RF=2, threshold=0.2) #pvalue=3.8352e-19
+ """T-test 2. Average contribution of different alpha when T=0.2"""
# base = repHist(30, Actions, alpha=1, RF=2, threshold=0.2)
# for alpha in np.arange(0.8, 1, 0.01):
# compare = repHist(30, Actions, alpha=alpha, RF=2, threshold=0.2)
# print("Alpha=", alpha, stats.ttest_ind(base, compare))
- """Epsilon-decrease 0.99 with 0.1 and 0.999"""
+
+
+ """T-test 3. Avg contribution of Epsilon-decrease 0.99 with 0.1 and 0.999"""
# base = repHist(30, Actions, multiArm='decrease', epsilon=0.99)
# for epsilon in (0.1, 0.999):
# compare = repHist(30, Actions, multiArm='decrease', epsilon=epsilon)
# print("Epsilon=", epsilon, stats.ttest_ind(base, compare))
- """T-TEST for 0.999-decrease 5000 iterations with 0.9"""
+ """T-test 4. Avg contribution of 0.999-decrease 5000 iterations with 0.9"""
# base = repHist(30, Actions, multiArm='decrease', epsilon=0.9)
# compare = repHist(30, Actions, multiArm='decrease', epsilon=0.999, I=5000)
# print(stats.ttest_ind(base, compare))
@@ -453,31 +372,5 @@ def main():
-
-
-
- """T-TEST K,P"""
-
- # t_test(30, Actions, K=(2, 99), P=0) #pvalue=0.4278
- # t_test(30, Actions, K=(2, 99), P=0.9) #pvalue=0.4541
- # for _ in range(5):
- # t_test(30, Actions, K=(2, 99), P=0.8) #pvalue=0.01502 ***
- # t_test(30, Actions, K=(2, 99), P=0.85) #pvalue=0.1931
- # t_test(30, Actions, K=(2, 99), P=0.75) #pvalue=0.5630
- # t_test(30, Actions, K=2, P=(0, 0.9)) #pvalue=0.9806
- # t_test(30, Actions, K=2, P=(0, 0.8)) #pvalue=0.4523
- # t_test(30, Actions, K=(2, 99), P=0.9) #pvalue=0.4541
- # t_test(30, Actions, K=(2, 99), P=0.7) #pvalue=0.3698
- # t_test(30, Actions, K=99, P=(0, 0.8)) #pvalue=0.8167
-
-
-
- # base = repHist(30, Actions, K=99)
- # for p in np.arange(0.76, 0.85, 0.01): #ALL >.05
- # compare = repHist(30, Actions, K=2, P=p)
- # print("K=2, P=", p, stats.ttest_ind(base, compare))
-
-
-
if __name__ == '__main__':
main()
\ No newline at end of file