def test_jointLocalMc1(self):
"""
Updates computed by hand given:
v = [45,20], l = 1
p(q_1 = 20) = 0.5
p(q_1 = 30) = 0.5
p(q_2 = 15) = 0.2
p(q_2 = 20) = 0.8
Under independent prices p(q_1,q_2) = p(q_1)p(q_2)
Hence:
p(q_1 = 20, q_2 = 15) = 0.1
p(q_1 = 20, q_2 = 20) = 0.4
p(q_1 = 30, q_2 = 15) = 0.1
p(q_1 = 30, q_2 = 20) = 0.4
The joint pdf is represented with 1000 samples.
100 samples of [20,15]
400 samples of [20,20]
100 samples of [30,15]
400 samples of [30,20]
Ground Truth Anwers:
Starting at [25,25]
1.1) b1 <- 25
1.2) b2 <- 10
2.1) b1 <- 45
2.2) b2 <- 0
3.1) b1 <- 25
3.2) b2 <- 0
Therefore, starting at [25,25] converges to [45,0] after 3 iterations
"""
samples = numpy.zeros((1000,2))
samples[:100,:] = numpy.asarray([20,15])
samples[100:500,:] = numpy.asarray([20,20])
samples[500:600,:] = numpy.asarray([30,15])
samples[600:,:] = numpy.asarray([30,20])
m=2
l = 1
v = [45,20]
bundles = listBundles(m)
revenue = msListRevenue(bundles, v, l)
bids = numpy.asarray([25.,25.])
bids[0] = jointLocalUpdateMc(bundles,revenue,bids,0,samples,True)
print bids
numpy.testing.assert_equal(bids[0], 25, "Update 1.1 Failed", True)
bids[1] = jointLocalUpdateMc(bundles,revenue,bids,1,samples,True)
print bids
numpy.testing.assert_equal(bids[1], 10, "Update 1.2 Failed", True )
bids[0] = jointLocalUpdateMc(bundles,revenue,bids,0,samples,True)
print bids
numpy.testing.assert_equal(bids[0], 45, "Update 2.1 Failed", True)
bids[1] = jointLocalUpdateMc(bundles,revenue,bids,1,samples,True)
print bids
numpy.testing.assert_equal(bids[1], 0, "Update 2.2 Failed", True)
bids[0] = jointLocalUpdateMc(bundles,revenue,bids,0,samples,True)
print bids
numpy.testing.assert_equal(bids[0], 45, "Update 3.1 Failed", True)
bids[1] = jointLocalUpdateMc(bundles,revenue,bids,1,samples,True)
print bids
numpy.testing.assert_equal(bids[1], 0, "Update 3.2 Failed", True)
bids = numpy.asarray([25.,25.])
bids, converged, itr, tol = jointLocalMc(bundles, revenue, bids, samples, verbose=True,ret='all')
print bids
print converged
print itr
print tol
def main():
desc="Compute jointLocal bids given v's, l's initial bids and parameters"
parser = argparse.ArgumentParser(description=desc)
parser.add_argument('-s','--samplesFile',dest='samplesFile',
required=True,type=str,
help='Samples (text file)')
parser.add_argument('-v','--vfile',dest='vfile',
required=True,type=str,
help='Valuation v vector txt file.')
parser.add_argument('-l','--lfile',dest='lfile',
required=True,type=str,
help='Valuation lambda txt file.')
parser.add_argument('-ib','--initBidFile',dest='initBidFile',
required=True,type=str,
help='Initial Bids txt file.')
parser.add_argument('-es','--evalSamples',dest='evalFile',
required=True,type=str,
help='Evaluation samples file.')
parser.add_argument('-n','--name',dest='name',
required=False,type=str,
default='jointLocalMc',
help='output file name prefix.')
parser.add_argument('-o','--odir',dest='odir',
required=True,type=str,
help='Output directory.')
parser.add_argument('-mi','--maxitr', dest='maxitr',
required=False, default=100, type=int,
help='Maximum Update Iterations')
parser.add_argument('-t','--tol',dest='tol',
required=False, default=1e-5,
type=float,help='L2 Update Stopping Tolerance')
parser.add_argument('--verbose',dest='verbose',
required=False,default=False,
type=bool,help='Output debugg information')
args = parser.parse_args()
args.odir = os.path.realpath(args.odir)
jointSamples = numpy.loadtxt(os.path.realpath(args.samplesFile))
initBids = numpy.loadtxt(os.path.realpath(args.initBidFile))
vmat = numpy.loadtxt(os.path.realpath(args.vfile))
lmat = numpy.loadtxt(os.path.realpath(args.lfile))
evalSamples = numpy.loadtxt(os.path.realpath(args.evalFile))
m = initBids.shape[1]
bundles = listBundles(m)
bids = numpy.zeros(initBids.shape)
es = numpy.zeros(initBids.shape[0])
for itr, initBid, v, l in zip(xrange(vmat.shape[0]),initBids,vmat,lmat):
print 'Iteration = {0}'.format(itr)
revenue = msListRevenue(bundles,v,l)
bids[itr,:], converged, nItr, d = \
jointLocalMc(bundles,revenue,initBid,jointSamples,
args.maxitr,args.tol,args.verbose,'all')
brd = {}
for b,r in zip(bundles,revenue):
brd[tuple(b)] = r
es[itr] = expectedSurplus_(brd, bids[itr,:], evalSamples)
print '\t bid = {0}'.format(bids[itr,:])
print '\t converged = {0}'.format(converged)
print '\t Num. Iterations = {0}'.format(nItr)
print '\t d = {0}'.format(d)
print '\t Expected Surplus = {0}'.format(es[itr])
numpy.savetxt(os.path.join(args.odir, args.name + '_bids.txt'), bids)
numpy.savetxt(os.path.join(args.odir, args.name + '_expectedSurplus.txt'),es)
with open(os.path.join(args.odir,args.name + '_stats.txt'),'w') as f:
print >> f, numpy.mean(es)
print >> f, numpy.var(es)