def test_efp_efm_naive_compute(measure, normed, event):
# compute the EFP corresponding to the "icecreamcone" graph in the usual way
EFP_correlator = ef.EFP([(0,1),(0,1),(0,2),(1,2)], measure=measure, normed=normed, coords='epxpypz', beta=2)(event)
# compute the EFP corresponding to "icecreamcone" as a contraction of EFMs with the metric
metric = np.diag([1,-1,-1,-1])
# compute several energy flow moments
EFM2 = ef.EFM(2, measure=measure, normed=normed, coords='epxpypz')(event)
EFM3 = ef.EFM(3, measure=measure, normed=normed, coords='epxpypz')(event)
EFP_contraction = np.einsum('abc,def,gh,ad,be,cg,fh->', EFM3, EFM3, EFM2, *([metric]*4))
# compute the EFP corresponding to "icecreamcone" with explicit sums
if measure == 'hadrefm':
zs = np.sqrt(event[:,1]**2 + event[:,2]**2)
if measure == 'eeefm':
zs = event[:,0]
ns = (event/zs[:,np.newaxis])
thetas = np.asarray([[np.sum(2*ni*nj*np.asarray([1,-1,-1,-1])) for ni in ns] for nj in ns])
if normed:
zs /= zs.sum()
EFP_sums = 0
for i in range(len(event)):
for j in range(len(event)):
for k in range(len(event)):
EFP_sums += zs[i] * zs[j] * zs[k] * thetas[i,j] * thetas[i,j] * thetas[i,k] * thetas[j,k]
# ensure that the two values agree
assert epsilon_percent(EFP_correlator, EFP_contraction, 10**-12)
assert epsilon_percent(EFP_correlator, EFP_sums, 10**-12)
def test_efm_batch_compute(v, M, measure, normed):
events = ef.gen_random_events(2, M)
e = ef.EFM(v, measure=measure, normed=normed, coords='epxpypz')
r1 = [e.compute(event) for event in events]
r2 = e.batch_compute(events)
assert epsilon_percent(r1, r2, 10**-1)
def test_efm_vs_efmset_compute(sigs, M, measure, normed):
efmset = ef.EFMSet(sigs, measure=measure, normed=normed, coords='epxpypz')
efms = [
ef.EFM(*sig, measure=measure, normed=normed, coords='epxpypz')
for sig in sigs
]
for event in ef.gen_random_events(2, M):
efm_dict = efmset.compute(event)
for sig, efm in zip(sigs, efms):
print(sig, np.max(np.abs(efm_dict[sig] - efm.compute(event))))
assert epsilon_percent(efm_dict[sig], efm.compute(event), 10**-10)
def test_efm_vs_efmset_batch_compute(sigs, M, measure, normed):
efmset = ef.EFMSet(sigs, measure=measure, normed=normed, coords='epxpypz')
efms = [
ef.EFM(*sig, measure=measure, normed=normed, coords='epxpypz')
for sig in sigs
]
events = ef.gen_random_events(2, M)
efm_dict = efmset.batch_compute(events)
for sig, efm in zip(sigs, efms):
results = efm.batch_compute(events)
for i in range(len(events)):
assert epsilon_percent(efm_dict[i][sig], results[i], 10**-10)
def test_efms(v, measure, normed, M):
events = ef.gen_random_events(2, M)
e = ef.EFM(v, measure=measure, normed=normed, coords='epxpypz')
for event in events:
if measure == 'hadrefm':
zs = np.atleast_1d(ef.pts_from_p4s(event))
elif measure == 'eeefm':
zs = event[:, 0]
nhats = event / zs[:, np.newaxis]
if normed:
zs = zs / zs.sum()
e_ans = e.compute(event)
if v == 0:
assert epsilon_percent(e_ans, zs.sum(), 10**-13)
else:
s_ans = slow_efm(zs, nhats, v)
assert epsilon_percent(s_ans, e_ans, 10**-13)