def test_F2_F1(self):
'''<d_id_j ψ|H|ψ>, <d_iψ|H|d_jψ>'''
Sx1, Sy1, Sz1 = N_body_spins(0.5, 1, 6)
Sx2, Sy2, Sz2 = N_body_spins(0.5, 2, 6)
Sx3, Sy3, Sz3 = N_body_spins(0.5, 3, 6)
Sx4, Sy4, Sz4 = N_body_spins(0.5, 4, 6)
Sx5, Sy5, Sz5 = N_body_spins(0.5, 5, 6)
Sx6, Sy6, Sz6 = N_body_spins(0.5, 6, 6)
Sx12, Sy12, Sz12 = N_body_spins(0.5, 1, 2)
Sx22, Sy22, Sz22 = N_body_spins(0.5, 2, 2)
mps = self.mps_0_6.left_canonicalise()
listH = [Sz12@Sz22+Sx12, Sz12@Sz22+Sx12+Sx22, Sz12@Sz22+Sx22+Sx12+Sx22, Sz12@Sz22+Sz12+Sx22, Sz12@Sz22+Sx22]
eyeH = [(1/(mps.L-1))*eye(4) for _ in range(5)]
fullH = Sz1@Sz2+Sz2@Sz3+Sz3@Sz4+Sz4@Sz5+Sz5@Sz6+Sx1+Sx2+Sx3+Sx4+Sx5+Sx6
for i, j in product(range(mps.L), range(mps.L)):
## F2 = <d_id_j ψ|H|ψ>
# zero for H = I
self.assertTrue(allclose(mps.F2(i, j, eyeH, testing=True), 0))
# Test gauge projectors are in the right place
mps.right_canonicalise()
l, r = mps.get_envs()
z1 = ncon([mps.F2(i, j, listH, testing=True), l(i-1)@c(mps[i])], [[1, 2, 3, -1, -2, -3], [1, 2, 3]])
z2 = ncon([mps.F2(i, j, listH, testing=True), l(j-1)@c(mps[j])], [[-1, -2, -3, 1, 2, 3], [1, 2, 3]])
self.assertTrue(allclose(z1, 0))
self.assertTrue(allclose(z2, 0))
mps.left_canonicalise()
l, r = mps.get_envs()
z1 = ncon([mps.F2(i, j, listH, testing=True), l(i-1)@c(mps[i])], [[1, 2, 3, -1, -2, -3], [1, 2, 3]])
z2 = ncon([mps.F2(i, j, listH, testing=True), l(j-1)@c(mps[j])], [[-1, -2, -3, 1, 2, 3], [1, 2, 3]])
self.assertTrue(allclose(z1, 0))
self.assertTrue(allclose(z2, 0))
## F1 = <d_iψ|H|d_jψ>
# For H = I: should be equal to δ_{ij} pr(i)
if i!=j:
self.assertTrue(allclose(mps.F1(i, j, eyeH, testing=True), 0))
if i==j:
b = mps.F1(i, j, eyeH, testing=True)
a = ncon([mps.left_null_projector(i), inv(r(i))], [[-1, -2, -4, -5], [-3, -6]])
self.assertTrue(allclose(a, b))
# Test gauge projectors are in the right place
mps.left_canonicalise()
l, r = mps.get_envs()
z1 = td(mps.F1(i, j, listH, testing=True), l(i-1)@c(mps[i]), [[0, 1, 2], [0, 1, 2]])
z1 = td(mps.F1(i, j, listH, testing=True), l(j-1)@mps[j], [[3, 4, 5], [0, 1, 2]])
self.assertTrue(allclose(z1, 0))
self.assertTrue(allclose(z2, 0))
mps.right_canonicalise()
l, r = mps.get_envs()
z1 = td(mps.F1(i, j, listH, testing=True), l(i-1)@c(mps[i]), [[0, 1, 2], [0, 1, 2]])
z1 = td(mps.F1(i, j, listH, testing=True), l(j-1)@mps[j], [[3, 4, 5], [0, 1, 2]])
self.assertTrue(allclose(z1, 0))
self.assertTrue(allclose(z2, 0))
def EJ(self):
ee = 2 * self.eeMatrix()
td = np.tensordot
mo = self.mo_vectors
out = td(mo, ee, axes=(1, 0))
out = td(mo, out, axes=(1, 1))
out = td(mo, out, axes=(1, 2))
out = td(mo, out, axes=(1, 3))
occ = int(np.sum(self.occupation) / 2.)
EJ = [out[a, a, b, b] for a in range(occ) for b in range(occ)]
return sum(EJ)
def EX(self):
ex = -np.swapaxes(self.eeMatrix(), 1, 2)
td = np.tensordot
mo = self.mo_vectors
out = td(mo, ex, axes=(1, 0))
out = td(mo, out, axes=(1, 1))
out = td(mo, out, axes=(1, 2))
out = td(mo, out, axes=(1, 3))
occ = int(np.sum(self.occupation) / 2.)
EX = [out[a, a, b, b] for a in range(occ) for b in range(occ)]
return sum(EX)
def EX(self): ex = -np.swapaxes(self.eeMatrix(), 1,2) td = np.tensordot mo = self.mo_vectors out = td(mo, ex, axes=(1, 0)) out = td(mo, out, axes=(1, 1)) out = td(mo, out, axes=(1, 2)) out = td(mo, out, axes=(1, 3)) occ = int(np.sum(self.occupation) / 2.) EX = [out[a,a,b,b] for a in range(occ) for b in range(occ)] return sum(EX)
def EJ(self): ee = 2*self.eeMatrix() td = np.tensordot mo = self.mo_vectors out = td(mo, ee, axes=(1, 0)) out = td(mo, out, axes=(1, 1)) out = td(mo, out, axes=(1, 2)) out = td(mo, out, axes=(1, 3)) occ = int(np.sum(self.occupation) / 2.) EJ = [out[a,a,b,b] for a in range(occ) for b in range(occ)] return sum(EJ)
def _dividing_total_data(self):
"""
分组函数，仅限内部一次使用
:return: 分组列表
"""
tmp1 = tmp2 = self.total_normal_data.index[0]
time_parts = []
part_series = []
self.text_out("正在生成处理数列")
# 将全部数据中分组为连续生产(生产间隔不大于DIVIDE_TIME)的若干大组
for date in self.total_normal_data.index:
if date - tmp1 > td(cf.divide_time, 'm'):
time_parts.append(self.total_normal_data[tmp2:date])
tmp2 = date
tmp1 = date
# 将每一大组中每SERIES_NUM的数据分为小组，将所有大组的每一小组添加到返回列表之中
for data_series in time_parts:
start_num, end_num = 0, cf.series_num
while end_num < len(data_series):
part_series.append(Series(data_series[start_num:end_num]))
start_num = end_num
end_num += cf.series_num
return part_series
def densityFitting(qmout, rho_basis=None):
psi_basis = qmout.basis
occ = [i for i in range(qmout.n_ao) if qmout.occupation[i] == 2][-1] + 1
D_matrix = densityMatrix(qmout)
if rho_basis is None:
rho_basis = psi_basis
psi_basis_data, psi_center, psi_lm = basisData(psi_basis)
rho_basis_data, rho_center, rho_lm = \
basisData(rho_basis, density=True)
NN = nnMatrix(rho_basis)
NE = neMatrix(psi_basis, rho_basis)
G = td(D_matrix, NE, axes=([0, 1], [1, 2]))
return np.linalg.solve(NN, G)
def densityFitting(qmout, rho_basis=None):
psi_basis = qmout.basis
occ = [i for i in range(qmout.n_ao)
if qmout.occupation[i]==2][-1] + 1
D_matrix = densityMatrix(qmout)
if rho_basis is None:
rho_basis = psi_basis
psi_basis_data, psi_center, psi_lm = basisData(psi_basis)
rho_basis_data, rho_center, rho_lm = \
basisData(rho_basis, density=True)
NN = nnMatrix(rho_basis)
NE = neMatrix(psi_basis, rho_basis)
G = td(D_matrix, NE, axes=([0,1], [1,2]))
return np.linalg.solve(NN, G)
last_day = pd.Timestamp('2017-06-05')
days = pd.date_range(first_day, last_day)
number_of_days = len(days)
sessions_list = []
columns = [
'date', 'session_start', 'session_end', 'session_length', 'longest_length',
'mode_changes_per_second', 'mode0', 'mode1', 'mode2', 'mode3', 'mode4',
'mode5', 'mode6', 'mode7', 'mode8'
]
for day in days:
date = day.date()
print('Preparing day', date)
start = day + td(9, 'h')
end = day + td(18, 'h')
points_filename = 'original data\%s.%s\max_values_%s-%s.txt' % (
day.day, day.month, day.day, day.month)
points = pd.read_csv(points_filename,
header=None,
names=['overpoints', 'timestamp'],
parse_dates=[1])
if points.empty:
continue
points['time'] = points.timestamp.apply(pd.Timestamp.time)
points = points[(points.time >= start.time())
& (points.time <= end.time())]
points['previous_timestamp'] = points.timestamp.shift()
points['difference'] = (points.timestamp -
mode_ids = {
'': 0,
'Lava': 1,
'Krajina': 2,
'VrstevniceFill': 3,
'Spad': 4,
'Voda': 5,
'Psychedelicky': 6,
'Vrstevnice': 7,
'Vyukovy': 8
}
for day in days:
print('Preparing day', day.date())
start = day + td(9, 'h')
end = day + td(18, 'h')
seconds = pd.date_range(start, end, freq='S')
mode_filename = '%s.%s\mode_change_%s-%s.txt' % (day.day, day.month,
day.day, day.month)
modes = pd.read_csv(mode_filename,
header=None,
names=['playmode', 'timestamp'],
parse_dates=[1])
modes['time'] = modes.timestamp.apply(pd.Timestamp.time)
modes['previous_timestamp'] = modes.timestamp.shift()
modes['previous_length'] = modes.timestamp - modes.previous_timestamp
modes['event'] = 'automatic'
modes.loc[modes.previous_length < td(1, 'm'), 'event'] = 'click'
import pandas as pd
from numpy import timedelta64 as td
import matplotlib.pyplot as plt
import numpy as np
from sklearn.linear_model import LogisticRegression, LinearRegression
import mord
minute = td(1, 'm')
video_end = pd.Timestamp('26-04-2017T16:00:00')
end = pd.Timestamp('26-04-2017T18:00:00')
modes = pd.read_csv('original data/26.4/mode_change_26-4.txt',
header=None,
names=['playmode', 'timestamp'],
parse_dates=[1])
modes['previous_timestamp'] = modes.timestamp.shift()
modes['previous_length'] = modes.timestamp - modes.previous_timestamp
modes['length'] = modes.timestamp.shift(-1) - modes.timestamp
modes['to_video_end'] = end - modes.timestamp
modes['event'] = 'automatic'
modes.loc[modes.previous_length < minute, 'event'] = 'click'
# print(modes[modes.to_video_end > td(0)].loc[:, ['playmode', 'previous_length', 'length']])
points = pd.read_csv('original data/26.4/max_values_26-4.txt',
header=None,
names=['overpoints', 'timestamp'],
parse_dates=[1])
points['previous_timestamp'] = points.timestamp.shift()
points['to_end'] = end - points.timestamp
