def revenueIronedMyersonAuctionFromSamples(trainingSamples,testSamples):
# trainingSamples = float[NUM_PLAYERS][NUM_TEST_SAMPLES]
'''generate many means and sigmas from trainingSamples'''
distributions=[getDistribution(trainingSet) for trainingSet in trainingSamples]
'''
test for exceptions
'''
'''testSamples are the new bids'''
#DO FOR ALL TEST SAMPLES
totalRevenue=0
revenues=[]
samples = zip(*testSamples)
numSamples = len(samples)
for i in xrange(numSamples):
bids = samples[i]
#call to regular ironedMyersonAuction since I know bids and distribution
virtualValues=[ironedVirtualValuation(bids[i], distributions[i]) for i in range(len(bids)) ]
reserves=[0]*len(bids)
(allocation, priceListAux) = eagerVickrey(virtualValues,reserves)
if sum(allocation) == 0:
priceList = priceListAux
else:
for i in range(len(allocation)):
if allocation[i]!=0: #in this case there is only one winner, with allocation set to 1
winner = i
secondVirtualValue = priceListAux[winner]
secondVirtualValue = max(secondVirtualValue,0)
price = scipy.optimize.fsolve(auxRootFindingIroned, bids[winner], (secondVirtualValue, distributions[winner], 0, bids[winner]))
priceList = [0 for i in allocation]
priceList[winner]=price
revenue = sum(priceList)
totalRevenue+=revenue
revenues.append(revenue)
return revenues
def revenueMyersonAuctionFromSamples(trainingSamples,testSamples, flagReserveInfo=False):
# trainingSamples = float[NUM_PLAYERS][NUM_TEST_SAMPLES]
'''generate many means and sigmas from trainingSamples'''
distributions=[getDistribution(trainingSet) for trainingSet in trainingSamples]
'''
test for exceptions
'''
'''testSamples are the new bids'''
#DO FOR ALL TEST SAMPLES
totalRevenue=0
revenues=[]
samples = zip(*testSamples)
numSamples = len(samples)
if (flagReserveInfo==False):
for i in xrange(numSamples):
bids = samples[i]
virtualValues=[virtualValuation(bids[i], distributions[i]) for i in range(len(bids)) ]
reserves=[0]*len(bids)
(allocation, priceListAux) = eagerVickrey(virtualValues,reserves)
if sum(allocation) == 0:
priceList = priceListAux
else:
for i in range(len(allocation)):
if allocation[i]!=0: #in this case there is only one winner, with allocation set to 1
winner = i
secondVirtualValue = priceListAux[winner]
secondVirtualValue = max(secondVirtualValue,0)
price = scipy.optimize.fsolve(auxRootFinding, bids[winner], (secondVirtualValue, distributions[winner], 0, bids[winner]))
priceList = [0 for i in allocation]
priceList[winner]=price[0]
revenue = sum(priceList)
totalRevenue+=revenue
revenues.append(revenue)
return revenues
else: #if (flagReserveInfo==True):
revenues1 = []
revenue1=0
counterNotSell1=0
counterPriceFromReserve1=0
avgReservePrice1 =0
for i in xrange(numSamples):
bids = samples[i]
(allocation,prices,isPriceFromReserve, reservePrice) = myersonAuction(bids, distributions, True)
if isPriceFromReserve==None:
counterNotSell1+=1
if isPriceFromReserve==True:
counterPriceFromReserve1+=1
avgReservePrice1=reservePrice
revenue1 = sum(prices)
revenues1.append(revenue1)
avgRevenue1 = sum(revenues1) / (numSamples * 1.0)
avgRevenue1*=complex(1,0)
avgRevenue1=avgRevenue1.real
avgSellingPrice1= avgRevenue1*numSamples/(numSamples-counterNotSell1)
timeItemNotSold1 = counterNotSell1*1.0/numSamples
timeReservePriceDeterminant1 = counterPriceFromReserve1*1.0/numSamples
timeCompetitionDeterminant1 = 1 - timeReservePriceDeterminant1 - timeItemNotSold1
s1 = "AvgRevenue\t" + str(avgRevenue1)+ "\tAvgSellingPrice\t" + str(avgSellingPrice1) + \
"\t%timeItemNotSold\t" + str(timeItemNotSold1) + "\t%timeReservePriceDeterminant\t"+ str(timeReservePriceDeterminant1) + \
"\t%timeCompetitionDeterminant\t" + str(timeCompetitionDeterminant1) + "\tavgReservePrice\t" + str(avgReservePrice1)
return s1
