def __init__(self, n_arms):
self.dim = auctionHouse.NB_CATEGORIES
self.arms = \
np.random.random_integers(0, auctionHouse.MAX_BID, size=(n_arms, self.dim))
self.graph = graph.Graph()
# bids of other advertisers
self.bids = np.random.randint(
0, auctionHouse.MAX_BID + 1,
(auctionHouse.NB_ADVERTISERS, auctionHouse.NB_CATEGORIES))
# the environment knows the ad quality of the ads
self.adQualitiesVector = np.clip(
np.random.normal(0.5, 0.1, auctionHouse.NB_ADVERTISERS), 0.1, 0.9)
# the environment also knows the click value for each advertiser
self.valuesOfClick = np.clip(
np.random.normal(2, 0.5, auctionHouse.NB_ADVERTISERS), 0.5, 5.0)
self.history = []
self.learningAdvertiserWonAuctions = []
# keep track of history of auctions
self.learningAdvertiserWonAuctionsHistory = []
# chances for each advertiser to change its bid at each time step
self.changeProbability = 0.1
import Q1.MonteCarlo as MonteCarlo
import matplotlib.pyplot as plt
# reward when a click happens
VALUE_OF_CLICK = 3
# prob to click on the ad knowing we looked at it
AD_QUALITY = 0.7
np.random.seed(0)
# Ad qualities of all advertisers (0 is the learning one)
AdQualitiesVector = np.clip(np.random.normal(0.5, 0.1, auctionHouse.NB_ADVERTISERS), 0.1, 0.9)
AdQualitiesVector[0] = AD_QUALITY
### GRAPH
graph = graph.Graph()
# randomize the bids, may be improved later
bids = np.random.randint(0, auctionHouse.MAX_BID + 1,
(auctionHouse.NB_ADVERTISERS, auctionHouse.NB_CATEGORIES))
rewardHistory = [] # to later draw curve of evolution of reward
rewardHistoryWithoutRollbacks = []
labels = [] # to annotate points
### BIDS/AUCTIONS
# sets bids of learning advertiser to 0
bids[0] = np.zeros(auctionHouse.NB_CATEGORIES) # set bids of current advertiser to 0
previousReward = 0 # when bids are 0, the reward will be 0
currentImprovedCategory = 0 # to jump to 0 at next iteration
import environment.graph as graph
import numpy as np
import time
g = graph.Graph()
np.random.seed(int(time.time()))
max_output = g.nbNodes
## Build a random connection matrix undirected
connection_matrix = g.generateConnectivityMatrix()
print(connection_matrix)
print(g.activation_probabilities_matrix)
g.changeTransitionProbabilities2([1, 1, 1, 1, 1],
g.activation_probabilities_matrix)
print(g.prob_matrix)
g.display()