def acc(self):
"""Compute the per-class accuracies and the overall accuracy.
Args:
scores (torch.FloatTensor, shape (B?, C, N):
raw scores for each class.
labels (torch.LongTensor, shape (B?, N)):
ground truth labels.
Returns:
A list of floats of length num_classes+1.
Consists of per class accuracy. Last item is Overall Accuracy.
"""
if self.confusion_matrix is None:
return None
accs = []
for label in range(self.num_classes):
tp = np.longlong(self.confusion_matrix[label, label])
fn = np.longlong(self.confusion_matrix[label, :].sum()) - tp
if tp + fn == 0:
acc = float('nan')
else:
acc = tp / (tp + fn)
accs.append(acc)
accs.append(np.nanmean(accs))
return accs
def getIouScoreForLabel(self, label):
# Calculate and return IOU score for a particular label (train_id)
if label == self.voidClass:
return float('nan')
# the number of true positive pixels for this label
# the entry on the diagonal of the confusion matrix
tp = np.longlong(self.confMatrix[label, label])
# the number of false negative pixels for this label
# the row sum of the matching row in the confusion matrix
# minus the diagonal entry
fn = np.longlong(self.confMatrix[label, :].sum()) - tp
# the number of false positive pixels for this labels
# Only pixels that are not on a pixel with ground truth label that is ignored
# The column sum of the corresponding column in the confusion matrix
# without the ignored rows and without the actual label of interest
notIgnored = [
l for l in self.validClasses
if not l == self.voidClass and not l == label
]
fp = np.longlong(self.confMatrix[notIgnored, label].sum())
# the denominator of the IOU score
denom = (tp + fp + fn)
if denom == 0:
return float('nan')
# return IOU
return float(tp) / denom
def getIouScoreForLabel(label, confMatrix, args):
if id2label[label].ignoreInEval:
return float('nan')
# the number of true positive pixels for this label
# the entry on the diagonal of the confusion matrix
tp = np.longlong(confMatrix[label, label])
# the number of false negative pixels for this label
# the row sum of the matching row in the confusion matrix
# minus the diagonal entry
fn = np.longlong(confMatrix[label, :].sum()) - tp
# the number of false positive pixels for this labels
# Only pixels that are not on a pixel with ground truth label that is ignored
# The column sum of the corresponding column in the confusion matrix
# without the ignored rows and without the actual label of interest
notIgnored = [
l for l in args.evalLabels
if not id2label[l].ignoreInEval and not l == label
]
fp = np.longlong(confMatrix[notIgnored, label].sum())
# the denominator of the IOU score
denom = (tp + fp + fn)
if denom == 0:
return float('nan')
# return IOU
return float(tp) / denom
def decode(code, a):
length = len(code)
decode = list(range(length))
decode[0] = np.longlong(code[0])
for i in range(length-1):
decode[i+1] = np.longlong(decode[i])*u +(code[i+1]-7.5)*a
return decode
def iou(self):
"""Compute the per-class IoU and the mean IoU.
Args:
scores (torch.FloatTensor, shape (B?, C, N):
raw scores for each class.
labels (torch.LongTensor, shape (B?, N)):
ground truth labels.
Returns:
A list of floats of length num_classes+1.
Consists of per class IoU. Last item is mIoU.
"""
if self.confusion_matrix is None:
return None
ious = []
for label in range(self.num_classes):
tp = np.longlong(self.confusion_matrix[label, label])
fn = np.longlong(self.confusion_matrix[label, :].sum()) - tp
fp = np.longlong(self.confusion_matrix[:, label].sum()) - tp
if tp + fp + fn == 0:
iou = float('nan')
else:
iou = (tp) / (tp + fp + fn)
ious.append(iou)
ious.append(np.nanmean(ious))
return ious
def decode(code, a):
length = len(code)
decode = list(range(length))
decode[0] = np.longlong(code[0])
for i in range(length-1):
#当前解码值等于上一次解码后的值加上压缩后的差值乘上量化因子
decode[i+1] = np.longlong(decode[i])*u +(code[i+1]-7.5)*a
return decode
def snr(self, data1, data2):
length = len(data1)
sum1 = np.longlong(0)
sum2 = np.longlong(0)
for i in range(length - 1):
n1 = np.longlong(data1[i])**2 / length # np.longlong防止计算时溢出
sum1 = sum1 + n1
n2 = np.longlong(data1[i] - data2[i])**2 / length
sum2 = sum2 + n2
return 10 * np.log10(sum1 / sum2)
def getIouScoreForLabel(label, cm):
tp = np.longlong(cm[label, label])
fn = np.longlong(cm[label, :].sum()) - tp
fp = np.longlong(cm[:, label].sum()) - tp
# the denominator of the IOU score
denom = (tp + fp + fn)
if denom == 0:
return float('nan')
# return IOU
return float(tp) / denom
def get_iou(label_id, confusion, num_classes):
tp = np.longlong(confusion[label_id, label_id])
fn = np.longlong(confusion[label_id, 1:].sum()) - tp
not_ignored = [l for l in range(1, num_classes) if not l == label_id]
fp = np.longlong(confusion[not_ignored, label_id].sum())
denom = (tp + fp + fn)
if denom == 0:
return (1, 1, 1, 1)
return (float(tp + 1e-6) / (denom + 1e-6),
float(tp + 1e-6) / float(tp + fn + 1e-6), tp, denom)
def get_acc(label_id, confusion):
if not label_id in VALID_CLASS_IDS:
return float('nan')
# #true positives
tp = np.longlong(confusion[label_id, label_id])
# #false negatives
fn = np.longlong(confusion[label_id, :].sum()) - tp
denom = (tp + fn)
if denom == 0:
return float('nan')
return (float(tp) / denom, tp, denom)
def get_iou(label_id, confusion):
if not label_id in VALID_CLASS_IDS:
return float('nan')
tp = np.longlong(confusion[label_id, label_id])
fn = np.longlong(confusion[label_id, 0:].sum()) - tp
not_ignored = [l for l in VALID_CLASS_IDS if not l == label_id]
fp = np.longlong(confusion[not_ignored, label_id].sum())
denom = (tp + fp + fn)
if denom == 0:
return (1, 1, 1, 1)
return (float(tp) / denom, float(tp) / float(tp + fn + 1e-6), tp, denom)
def get_iou(label_id, confusion):
# true positives
tp = np.longlong(confusion[label_id, label_id])
# false positives
fp = np.longlong(confusion[label_id, :].sum()) - tp
# false negatives
fn = np.longlong(confusion[:, label_id].sum()) - tp
denom = (tp + fp + fn)
if denom == 0:
return (0, 0, 0)
return (float(tp) / denom, tp, denom)
def get_iou(label_id, confusion):
# true positives
tp = np.longlong(confusion[label_id, label_id])
# false negatives
fn = np.longlong(confusion[label_id, :].sum()) - tp
# false positives
not_ignored = [l for l in data.VALID_CLASS_IDS if not l == label_id]
fp = np.longlong(confusion[not_ignored, label_id].sum())
denom = (tp + fp + fn)
if denom == 0:
return float('nan')
return (float(tp) / denom, tp, denom)
def snr(f):
cn, wlen, sc = DPCM_encode(f)
dc = DPCM_decode(f)
s = 0
n = 0
for a in range(len(sc)):
af = numpy.longlong(dc[a])
bef = numpy.longlong(sc[a])
p = af * af
q = (af - bef) * (af - bef)
s = s + p
n = n + q
return numpy.log10(s / n) * 10
def get_class_scores(self):
scores = []
for i in range(self.class_num):
tp = np.longlong(self.conf_mat[i, i])
gti = np.longlong(self.conf_mat[i, :].sum())
resi = np.longlong(self.conf_mat[:, i].sum())
denom = gti + resi - tp
try:
res = float(tp) / denom
except ZeroDivisionError:
res = 0
scores.append(res)
return scores
def get_iou(label_id, confusion):
# true positives
tp = np.longlong(confusion[label_id, label_id])
# false negatives
fn = np.longlong(confusion[label_id, :].sum()) - tp
# false positives
not_ignored = [l for l in range(20) if not l == label_id]
fp = np.longlong(confusion[not_ignored, label_id].sum())
denom = (tp + fp + fn)
if denom == 0:
return False
return (float(tp) / denom, tp, denom)
def energy(waveData):
wlen = len(waveData)
step = 256 # 每帧采样点数
frameNum = math.ceil(wlen / step) #函数返回数字的上入整数,帧数
ener = []
for i in range(frameNum):
curFrame = waveData[np.arange(i * step, min(i * step + step, wlen))]
sum = np.longlong(0) #防止溢出
for j in range(len(curFrame)):
n = np.longlong(curFrame[j])**2 #np.longlong防止计算时溢出
sum = sum + n
ener.append(sum)
return ener
def PhaseKai2opt(k_vector, noisy_original_image_fft, system_otf):
w = np.shape(noisy_original_image_fft)[0]
wo = np.int(w / 2)
noisy_original_image_fft = noisy_original_image_fft * (
1 - 1 * system_otf**10) #Increase the contrast by denoising
noisy_original_image_fft = noisy_original_image_fft * np.conj(
system_otf
) #Build term for minimisation (highlights places where i term is similar)
otf_cutoff = otf_smooth(system_otf)
DoubleMatSize = 0
if (
2 * otf_cutoff > wo
): #Contingency for the situtation where the size of the FFT is not large enough to fit in extra frequency info from SIM reconstruction
DoubleMatSize = 1
if (DoubleMatSize > 0):
t = 2 * w
noisy_original_image_fft_temp = np.zeros((t, t))
noisy_original_image_fft_temp[wo:w + wo,
wo:w + wo] = noisy_original_image_fft
noisy_original_image_fft = noisy_original_image_fft_temp
else:
t = w
to = np.int(t / 2)
u = np.linspace(0, t - 1, t)
v = np.linspace(0, t - 1, t)
[U, V] = np.meshgrid(u, v)
# Build term for comparison in cross-correlation (image with frequency added to it)
noisy_image_freqadd = np.exp(
-1j * 2 * np.pi * (k_vector[1] / t * (U - to)) +
(k_vector[0] / t * (V - to))) * fft.ifft2(noisy_original_image_fft)
noisy_image_freqadd_fft = fft.fft2(noisy_image_freqadd)
mA = np.longlong(
np.sum(noisy_original_image_fft * np.conj(noisy_image_freqadd_fft))
) # Sum across pixels of product of image with complex conjugate with frequency introduced.
mA = mA / np.longlong(
(np.sum(noisy_image_freqadd_fft * np.conj(noisy_image_freqadd_fft))
)) # Normalising cross-correlation term
#print(type(mA))
#print(-np.abs(mA))
correlation_FOM = -abs(
mA
) # Negative absolute value allows for minimisation; FOM = figure of merit
return (correlation_FOM)
def get_iou(label_id, confusion):
# true positives
tp = np.longlong(confusion[label_id, label_id])
# false negatives
fn = np.longlong(confusion[label_id, :].sum()) - tp
# false positives
not_ignored = [l for l in VALID_CLASS_IDS if not l == label_id]
fp = np.longlong(confusion[not_ignored, label_id].sum())
denom = (tp + fp + fn)
if denom == 0: #I guess that happens if a class is not represented in the training data
#return float('nan')
return (0, 0, 0)
return (float(tp) / denom, tp, denom)
def cal_snr(before, after):
x = np.longlong(0)
y = np.longlong(0)
length = len(before)
for i in range(length):
# print('before:'+str(before[i]))
# print('after:'+str(after[i]))
x += before[i]**2 / length # 防止溢出
y += (before[i] - after[i])**2 / length # 防止溢出
# print('x:'+str(x))
# print('y:'+str(y))
# print(x)
# print(y)
return 10 * np.log10(x / y)
def get_iou(label_id, confusion):
if not label_id in VALID_INSTANCE_ID:
return (float('nan'), 0, 0)
# #true positives
tp = np.longlong(confusion[label_id, label_id])
# #false negatives
fn = np.longlong(confusion[label_id, :].sum()) - tp
# #false positives
not_ignored = [l for l in VALID_INSTANCE_ID if not l == label_id]
fp = np.longlong(confusion[not_ignored, label_id].sum())
denom = (tp + fp + fn)
if denom == 0:
return (float('nan'), 0, 0)
return (float(tp) / denom, tp, denom)
def eval_ious(mat):
ious = np.zeros(3)
for l in range(2):
tp = np.longlong(mat[l, l])
fn = np.longlong(mat[l, :].sum()) - tp
notIgnored = [i for i in range(2) if not i == l]
fp = np.longlong(mat[notIgnored, l].sum())
denom = (tp + fp + fn)
if denom == 0:
print('error: denom is 0')
ious[l] = float(tp) / denom
return ious[:-1]
def generalized_eval_ious(mat):
n = mat.shape[0]
ious = np.zeros(n)
for l in range(n):
tp = np.longlong(mat[l, l])
fn = np.longlong(mat[l, :].sum()) - tp
notIgnored = [i for i in range(n) if not i == l]
fp = np.longlong(mat[notIgnored, l].sum())
denom = (tp + fp + fn)
if denom == 0:
print('error: denom is 0')
ious[l] = float(tp) / denom
return ious
def __init__(self):
self.tally_values = np.zeros([1])
self.unc_values = np.zeros([1])
self.xb = np.zeros([2])
self.yb = np.zeros([2])
self.zb = np.zeros([2])
self.nps = np.longlong(0)
def test_make_info():
"""Test some create_info properties."""
n_ch = np.longlong(1)
info = create_info(n_ch, 1000., 'eeg')
assert set(info.keys()) == set(RAW_INFO_FIELDS)
coil_types = {ch['coil_type'] for ch in info['chs']}
assert FIFF.FIFFV_COIL_EEG in coil_types
pytest.raises(TypeError, create_info, ch_names='Test Ch', sfreq=1000)
pytest.raises(ValueError, create_info, ch_names=['Test Ch'], sfreq=-1000)
pytest.raises(ValueError, create_info, ch_names=['Test Ch'], sfreq=1000,
ch_types=['eeg', 'eeg'])
pytest.raises(TypeError, create_info, ch_names=[np.array([1])],
sfreq=1000)
pytest.raises(KeyError, create_info, ch_names=['Test Ch'], sfreq=1000,
ch_types=np.array([1]))
pytest.raises(KeyError, create_info, ch_names=['Test Ch'], sfreq=1000,
ch_types='awesome')
pytest.raises(TypeError, create_info, ['Test Ch'], sfreq=1000,
montage=np.array([1]))
m = make_standard_montage('biosemi32')
info = create_info(ch_names=m.ch_names, sfreq=1000., ch_types='eeg')
info.set_montage(m)
ch_pos = [ch['loc'][:3] for ch in info['chs']]
ch_pos_mon = m._get_ch_pos()
ch_pos_mon = np.array(
[ch_pos_mon[ch_name] for ch_name in info['ch_names']])
# transform to head
ch_pos_mon += (0., 0., 0.04014)
assert_allclose(ch_pos, ch_pos_mon, atol=1e-5)
def p1(z, x):
k = 0
s = np.longlong(0.0)
for i in z:
s += i * pow(x, k)
k += 1
def make_Ranging_Code_Table(self):
# --- Find number of samples per spreading code ----------------------------
samples_Per_Code = self.samples_Per_Code
# --- Prepare the output matrix to speed up function -----------------------
Ranging_Code_Table = np.zeros((37, samples_Per_Code))
# --- Find time constants --------------------------------------------------
ts = 1.0 / self.sampling_Freq
tc = 1.0 / self.code_Freq_Basis
# === For all satellite PRN-s ...
for PRN in range(37):
# --- Generate Ranging code for given PRN -----------------------------------
Ranging_Code = self.generate_Ranging_Code(PRN)
# --- Make index array to read Ranging code values -------------------------
code_Value_Index = np.floor(ts * np.arange(0, samples_Per_Code) /
tc)
code_Value_Index = np.longlong(code_Value_Index)
code_Value_Index[-1] = 2045 ## is equal to 2045
Ranging_Code_Table[PRN] = Ranging_Code[code_Value_Index]
return Ranging_Code_Table
def rightNum(testX, testy, w, v):
pred = np.dot(w, testX.T).T + np.longlong(
np.sum((np.dot(testX, v)**2 - np.dot(testX**2, v**2)), axis=1).reshape(
len(testX), 1)) / 2.0
pred[pred >= 0] = 1
pred[pred < 0] = 0
return np.sum(pred == testy)
def test_numpy(self):
"""NumPy objects get serialized to readable JSON."""
l = [
np.float32(12.5),
np.float64(2.0),
np.float16(0.5),
np.bool(True),
np.bool(False),
np.bool_(True),
np.unicode_("hello"),
np.byte(12),
np.short(12),
np.intc(-13),
np.int_(0),
np.longlong(100),
np.intp(7),
np.ubyte(12),
np.ushort(12),
np.uintc(13),
np.ulonglong(100),
np.uintp(7),
np.int8(1),
np.int16(3),
np.int32(4),
np.int64(5),
np.uint8(1),
np.uint16(3),
np.uint32(4),
np.uint64(5),
]
l2 = [l, np.array([1, 2, 3])]
roundtripped = loads(dumps(l2, cls=EliotJSONEncoder))
self.assertEqual([l, [1, 2, 3]], roundtripped)
def test_numpy(self):
"""NumPy objects get serialized to readable JSON."""
l = [
np.float32(12.5),
np.float64(2.0),
np.float16(0.5),
np.bool(True),
np.bool(False),
np.bool_(True),
np.unicode_("hello"),
np.byte(12),
np.short(12),
np.intc(-13),
np.int_(0),
np.longlong(100),
np.intp(7),
np.ubyte(12),
np.ushort(12),
np.uintc(13),
np.ulonglong(100),
np.uintp(7),
np.int8(1),
np.int16(3),
np.int32(4),
np.int64(5),
np.uint8(1),
np.uint16(3),
np.uint32(4),
np.uint64(5),
]
l2 = [l, np.array([1, 2, 3])]
roundtripped = loads(dumps(l2, cls=EliotJSONEncoder))
self.assertEqual([l, [1, 2, 3]], roundtripped)
def decode(self, code, a, flag):
length = len(code)
decode = list(range(length))
decode[0] = np.longlong(code[0])
# 量化因子法
if flag == 0:
for i in range(length - 1):
decode[i + 1] = np.longlong(decode[i]) + (code[i + 1] - 8) * a
else:
for i in range(length - 1):
if code[i + 1] >= 8: # 负数
decode[i + 1] = decode[i] - math.exp(code[i + 1] -
8) + 1 # 解码，用于反馈
else:
decode[i + 1] = decode[i] + math.exp(
code[i + 1]) - 1 # 解码，用于反馈
return decode
def history_to_json(fname, snapshot, id):
"""Generate json version of merger history."""
id = np.longlong(id)
meraxes.set_little_h(fname)
gals = meraxes.read_gals(fname, snapshot, props=PROPS, quiet=True)
snaplist, _, _ = meraxes.read_snaplist(fname)
# get the ind of our start galaxy
ind = np.where(gals['ID'] == id)[0]
assert(len(ind) == 1)
ind = ind[0]
# read in the walk indices
fp_ind = deque()
np_ind = deque()
for snap in tqdm(snaplist[:snapshot+1], desc="Reading indices"):
try:
fp_ind.append(meraxes.read_firstprogenitor_indices(fname, snap))
except:
fp_ind.append([])
try:
np_ind.append(meraxes.read_nextprogenitor_indices(fname, snap))
except:
np_ind.append([])
# generate the graph (populates global variable `tree`)
walk(snapshot, ind, fp_ind, np_ind)
# attach the galaxies to the graph
for snap in tqdm(snaplist[:snapshot+1], desc="Generating graph"):
try:
gal = meraxes.read_gals(fname, snapshot=snap, quiet=True,
props=PROPS)
except IndexError:
continue
for nid in tree.nodes_iter():
node = tree.node[nid]
gal_snap, gal_ind = [int(v) for v in nid.split('|')]
if gal_snap == snap:
add_galaxy_to_node(node, gal[gal_ind])
# dump the tree to json
data = json_graph.tree_data(tree, root=node_id(snapshot, ind),
attrs=dict(id='name', children='children'))
fname_out = "tree_%09d.json" % id
with open(fname_out, "wb") as fd:
json.dump(data, fd)
print("Conversion complete: %s" % fname_out)
def test_make_info():
"""Test some create_info properties."""
n_ch = np.longlong(1)
info = create_info(n_ch, 1000., 'eeg')
assert set(info.keys()) == set(RAW_INFO_FIELDS)
coil_types = set([ch['coil_type'] for ch in info['chs']])
assert FIFF.FIFFV_COIL_EEG in coil_types
pytest.raises(TypeError, create_info, ch_names='Test Ch', sfreq=1000)
pytest.raises(ValueError, create_info, ch_names=['Test Ch'], sfreq=-1000)
pytest.raises(ValueError, create_info, ch_names=['Test Ch'], sfreq=1000,
ch_types=['eeg', 'eeg'])
pytest.raises(TypeError, create_info, ch_names=[np.array([1])],
sfreq=1000)
pytest.raises(TypeError, create_info, ch_names=['Test Ch'], sfreq=1000,
ch_types=np.array([1]))
pytest.raises(KeyError, create_info, ch_names=['Test Ch'], sfreq=1000,
ch_types='awesome')
pytest.raises(TypeError, create_info, ['Test Ch'], sfreq=1000,
ch_types=None, montage=np.array([1]))
m = read_montage('biosemi32')
info = create_info(ch_names=m.ch_names, sfreq=1000., ch_types='eeg',
montage=m)
ch_pos = [ch['loc'][:3] for ch in info['chs']]
assert_array_equal(ch_pos, m.pos)
names = ['nasion', 'lpa', 'rpa', '1', '2', '3', '4', '5']
d = read_dig_montage(hsp_fname, None, elp_fname, names, unit='m',
transform=False)
info = create_info(ch_names=m.ch_names, sfreq=1000., ch_types='eeg',
montage=d)
idents = [p['ident'] for p in info['dig']]
assert FIFF.FIFFV_POINT_NASION in idents
info = create_info(ch_names=m.ch_names, sfreq=1000., ch_types='eeg',
montage=[d, m])
ch_pos = [ch['loc'][:3] for ch in info['chs']]
assert_array_equal(ch_pos, m.pos)
idents = [p['ident'] for p in info['dig']]
assert (FIFF.FIFFV_POINT_NASION in idents)
info = create_info(ch_names=m.ch_names, sfreq=1000., ch_types='eeg',
montage=[d, 'biosemi32'])
ch_pos = [ch['loc'][:3] for ch in info['chs']]
assert_array_equal(ch_pos, m.pos)
idents = [p['ident'] for p in info['dig']]
assert (FIFF.FIFFV_POINT_NASION in idents)
assert info['meas_date'] is None
def Import_MCNPX_output(filename,
DOSE_FACTOR=1.0,
MESH_NUM=1,
VERBOSE=0):
"""
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;+
;AUTHOR: Justin Mikell, [email protected], justin[Dot][email protected]
;NAME: Import_MCNPX_output
;PURPOSE: to import a mesh tally (1D,2D, or 3D) from an MCNPX output file
;
;INPUTS:
;
;CATEGORY: Monte Carlo, MCNPX
;CALLING SEQUENCE: resultStruct = Import_MCNPX_output(filename [,/VERBOSE])
;KEYWORDS:
; DOSE_FACTOR: if set will multiply the tally data by this factor (default value of 1.0)
; /POS_VOLUME_OBJ: if set will return a pos volume object
; /VERBOSE: if set will print out information to console
;
;RESTRICTIONS:
;EXAMPLE:
;-
;MODIFICATION HISTORY:
;Created on 8-06-2010 by Justin Mikell
;;Added removal of extra white space 8-19-2010 JM
;;Added number of particles simulated to return struct 8-28-2010 JM
;TODO:
; -add support for multiple meshes
; -add support for 1D
; -add support 2D
;
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; docformat = 'rst'
;+
; :Author: Justin Mikell
; justin[Dot]mikell(AT]gmail.com
;
; :Description:
; Describe the procedure.
;
; :Uses:
; List the procedures/functions that this method calls here.
;
; :Params:
; filename: in, optional, type=string
; filename represents an MCNPX mesh tally outputfile with the following
; format::
;
; line 1: comment
; line 2: number_of_meshes number_of_particles_simulated
; line 3: ignore particle_type?
; line 4: 3rd integer is number of x boundaries (nx+1)
; 4th integer is number of y boundaries (ny+1)
; 5th integer is number of z boundaries (nz+1)
; line 5: ignore energy_range?
; line 6: ignore
; (repeat 3-6 for number of meshes)
; line 8: all the xb's are on this line separated by white space (in cm)
; line 9: all the yb's are on this line separated by white space (in cm)
; line 10: all the zb's are on this line separated by white space (in cm) (could also be angles depending on mesh)
; line 11 to line (ny*nz+10): nx*ny*nz values of output in MeV/(g*photon)
; each column represents x, each line represents y, ny lines implies change in z and reset y.
; line (ny*nz+11) to line (2*ny*nz+10): nx*ny*nz values of fractional uncertainty
;
; :Returns: {Tally_xyz:values,Unc_xyz:values, xb:xb,yb:yb,zb:zb, nps:nps}
; Tally_xyz is nx*ny*nz elements and represents the data stored at
; mesh(i,j,k).
; Unc_xyz is nx*ny*nz elements and represents the fractional uncertainty
; at mesh(i,j,k). (e.g. 0.01 means 1%)
; xb,yb,zb represent the voxel boundaries. (in cm)
; (The voxels may not be evenly spaced or the same size.)
; nps is the number of particles simulated.
;
; :Categories:
;
; :Keywords:
; DOSE_FACTOR
; POS_VOLUME_OBJ
; MESH_NUM
; VERBOSE
;
; :Examples:
; Please put an example::
;
; ;example IDL code
; ; output success
;
; regular comment again
;
; :History:
; Created by Justin Mikell on Sep 21, 2010
Ported to python by Justin Mikell on July 21, 2016
No POS_VOLUME_OBJ support yet.
"""
f = open(filename, 'r')
data = f.read()
f.close()
mylines = data.splitlines()
n_lines = len(mylines)
print("Importing data from :{0}".format(filename))
if(VERBOSE > 0):
print("first lines of file preceding data are...")
for i in np.arange(9):
print("{0}:{1}".format(i,mylines[i]))
print("Removing extra whitespace..")
for i in np.arange(len(mylines)):
mylines[i] = " ".join(mylines[i].split())
if(VERBOSE > 0):
print("first lines after removing extra whitespace...")
for i in np.arange(9):
print("{0}:'{1}'".format(i,mylines[i]))
#get the number of particles simulated
npsline = mylines[1].split()
nps = np.double(npsline[1])
total_meshes = np.uint(npsline[0])
if (MESH_NUM > total_meshes):
print("Warning mesh_num: {0} not found. Setting mesh_num=1".format(MESH_NUM))
MESH_NUM=1 #1=-based counting for mesh
if(VERBOSE > 0):
print("nps: {0}".format(nps))
print("total_meshes: {0}".format(total_meshes))
#read in all the mesh dimensions
mesh_dimensions_dict = {}
for i in np.arange(total_meshes):
boundsline = mylines[np.int(3+4*i)].split()
if(VERBOSE > 0):
print("reading bounds for mesh {0}".format(i))
print("boundsline:{0}".format(boundsline))
nx = np.long(boundsline[2])-1
ny = np.long(boundsline[3])-1
nz = np.long(boundsline[4])-1
#this will help with 1D and 2D arrays
if (nx == 0):
nx = nx+1
if (ny == 0):
ny = ny+1
if (nz == 0):
nz = nz+1
mesh_dimensions_dict[i] = [nx,ny,nz]
#Each mesh consists of 3+2*ny*nz lines
# 3 comes from xb,yb,zb
# 2*nx*ny comes from the dose values and their uncertainties
first_xb_line = 7+4*(total_meshes-1) #this is past all the "header" info
xb_line = np.long(first_xb_line)
print("xb_line:{0}".format(xb_line))
for i in np.arange(total_meshes-1):
xb_line = np.long(xb_line + 3 + 2*mesh_dimensions_dict[i][1]*mesh_dimensions_dict[i][2])
print("xb_line:{0}".format(xb_line))
#read in xbounds, ybounds, zbounds
xb = np.array(mylines[xb_line].split(), dtype=np.float)
yb = np.array(mylines[xb_line+1].split(), dtype=np.float)
zb = np.array(mylines[xb_line+2].split(), dtype=np.float)
#now set nx,ny,nz based on the mesh we have selected to read
nx = mesh_dimensions_dict[MESH_NUM-1][0]
ny = mesh_dimensions_dict[MESH_NUM-1][1]
nz = mesh_dimensions_dict[MESH_NUM-1][2]
#see if ny,nz agree with rest of file
if (np.longlong(2)*ny*nz) != (n_lines-10):
print("This may be a spherical/cylindrical mesh with an implicit 0.")
if(yb[ny] == 180):
print("This appears to be a spherical/cylindrical mesh file")
ans = 'NA'
while ( ans.upper() != 'Y' and ans.upper() != 'N'):
print("Treat as spherical/cylindrical with implicit 0 in polar angle? [y/n]")
ans = input('[y/n]')
if ans.upper() == 'Y':
yb = np.append(0, yb)
ny = ny + 1
#create the funcrz
print("reading in dose values now...")
tally_xyz = np.ndarray([nx,ny,nz], dtype=np.double)
for k in np.arange(nz):
for j in np.arange(ny):
temp = np.array(mylines[np.long(ny*k+j+(xb_line+3))].split(), dtype=np.double) #data starts 3 lines after xb
if(VERBOSE > 0):
print("Line number: {0}".format(np.long(ny*k+j+(xb_line+4))))
print("total elements on line: {0}".format(len(temp)))
if(len(temp) != nx):
print("number of imported elements doesnt match expected!!!")
exit(1)
tally_xyz[:,j,k] = temp
#now read the uncertainty
print("reading in uncertainty values...")
unc_xyz = np.ndarray([nx,ny,nz], dtype=np.double)
for k in np.arange(nz):
for j in np.arange(ny):
temp = np.array(mylines[np.long(ny*k+j+ny*nz+(xb_line+3))].split(), dtype=np.double)
if(VERBOSE > 0):
print("Line number: {0}".format(np.long(ny*k+j+ny*nz+(xb_line+4))))
print("total elements on line: {0}".format(len(temp)))
if(len(temp) != nx):
print("number of imported elements doesnt match expected!!!")
exit(1)
unc_xyz[:,j,k] = temp
tally_xyz = tally_xyz*DOSE_FACTOR
return({'tally_xyz':tally_xyz, 'unc_xyz':unc_xyz, 'xb':xb, 'yb':yb, 'zb':zb, 'nps':nps})
def import_from_mdata_ascii(self, filename):
""" Example
o = jkcm_mcnpx_rmesh()
filename = "/home/justin/001m"
o.import_from_mdata_ascii(filename)
"""
f = open(filename, 'r')
data = f.read()
f.close()
mylines = data.splitlines()
#get nps associated with mdata
self.nps = np.longlong(mylines[0].split()[5])
#find the nx,ny,nz line
p = re.compile('^f .*')
j=0
for i in np.arange(len(mylines)):
if(p.match(mylines[i])):
j=i
break
q = mylines[j].split()
nxyz = np.longlong(q[1])
nx = np.long(q[3])
ny = np.long(q[4])
nz = np.long(q[5])
#allocate your arrays
self.tally_values= np.zeros([nxyz])
self.unc_values=np.zeros([nxyz])
self.xb = np.zeros([nx+1])
self.yb = np.zeros([ny+1])
self.zb = np.zeros([nz+1])
#read in xbounds
t=0
for i in np.arange((j+1), len(mylines)):
t= t + len(mylines[i].split())
if(t == (nx+1)):
break
temp=''
for k in np.arange((j+1),i+1):
temp += mylines[k]
self.xb = np.array(temp.split(), dtype=np.double)
print("min/max xb: {0},{1}".format(min(self.xb),max(self.xb)))
#read in ybounds
j=i
t=0
for i in np.arange((j+1), len(mylines)):
t= t + len(mylines[i].split())
if(t == (ny+1)):
break
temp=''
for k in np.arange((j+1),i+1):
temp += mylines[k]
self.yb = np.array(temp.split(), dtype=np.double)
print("min/max yb: {0},{1}".format(min(self.yb),max(self.yb)))
#read in zbounds
j=i
t=0
for i in np.arange((j+1), len(mylines)):
t= t + len(mylines[i].split())
if(t == (nz+1)):
break
temp=''
for k in np.arange((j+1),i+1):
temp += mylines[k]
self.zb = np.array(temp.split(), dtype=np.double)
print("min/max zb: {0},{1}".format(min(self.zb),max(self.zb)))
#advance to tally values
p = re.compile('^vals.*')
j=0
for i in np.arange(len(mylines)):
if(p.match(mylines[i])):
j=i
break
#read everything into an array
tempArr = np.zeros([2*nxyz])
si=0
fi=0
for i in np.arange((j+1),len(mylines)):
temp = np.asarray(mylines[i].split(), dtype=np.double)
fi = si + len(temp)
tempArr[si:fi] = temp
si = fi
#separate absobred dose and uncertainty
self.unc_values = tempArr[1::2].reshape([nx,ny,nz],order='F')
self.tally_values = tempArr[0::2].reshape([nx,ny,nz], order='F')
#xc = 0.5*self.xb[0:-1] + 0.5*self.xb[1:]
#e= np.reshape(np.repeat(xc,nz),[nx,nz])
#e1=np.reshape(np.repeat(zc,nx), [nx,nz], order='F')
pmc_to_longlong = _PMCIntegers.pmc_to_longlong
def longlong_to_pmc(*args):
return _PMCIntegers.longlong_to_pmc(*args)
longlong_to_pmc = _PMCIntegers.longlong_to_pmc
try:
import numpy
RegisterPy2PMC(
is_py = lambda x: type(x) is numpy.longlong,
py2pmc = lambda x: longlong_to_pmc(long(x)),
)
RegisterPMC2Py(
is_pmc = pmc_is_longlong,
pmc2py = lambda x: numpy.longlong(pmc_to_longlong(x)),
)
except ImportError: pass
import ctypes
RegisterPy2PMC(
is_py = lambda x: type(x) is ctypes.c_longlong,
py2pmc = lambda x: longlong_to_pmc(x.value),)
RegisterPMC2Py(
is_pmc = pmc_is_longlong,
pmc2py = lambda x: ctypes.c_longlong(pmc_to_longlong(x)),
)
def generate_bispectrum_matrix2(self,n=5,n_guess_bsp=1e6,verbose=False,bsp_mat='sparse'):
''' Calculates the matrix to convert from uv phases to bispectra.
This version iterates through the sampling points in a vectorized way.
It saves all of the triangles, then removes the duplicates every 'n'
iterations. Reduce the number to save on ram but make it much slower.
n_guess_bsp: guess the number of bispectra and pre-allocate the memory
to save millions of 'append' calls (which was the slowest part). It must
be large enough to contain all of the bispectra, or you will get an error.
'''
nbsp=self.nbuv*(self.nbuv-1)*(self.nbuv-2) / 6
uv_to_bsp = np.zeros((n_guess_bsp,self.nbuv),dtype=np.long)
bsp_u = np.zeros((n_guess_bsp,3)) # the u points of each bispectrum point
bsp_v = np.zeros((n_guess_bsp,3)) # the v points of each bispectrum point
already_done = np.zeros((n_guess_bsp),dtype=np.longlong) # to track the sets of uv points that have already been counted
bsp_ix=0
uvrel=self.uvrel+np.transpose(self.uvrel)+1
nbits=np.longlong(np.ceil(np.log(self.nbuv)/np.log(10)))
print 'Calculating bispectrum matrix. Will take a few minutes.'
# Loop over the first pupil sampling point
tstart=time.time()
for ix1 in range(self.nbh):
# Loop over the second pupil sampling point
for ix2 in range(ix1+1,self.nbh):
# Rather than a for loop, vectorize it!
ix3s=np.arange(ix2+1,self.nbh)
n_ix3s=ix3s.size
if (bsp_ix+n_ix3s) > n_guess_bsp:
raise IndexError('Number of calculated bispectra exceeds the initial guess for the matrix size!')
# Find the baseline indices
b1_ix=uvrel[ix1,ix2]
b2_ixs=uvrel[ix2,ix3s]
b3_ixs=uvrel[ix1,ix3s] # we actually want the negative of this baseline
b1_ixs=np.repeat(b1_ix,n_ix3s)
# What uv points are these?
uv1=self.uv[b1_ixs,:]
uv2=self.uv[b2_ixs,:]
uv3=self.uv[b3_ixs,:]
# Are they already in the array? (any permutation of these baselines is the same)
# Convert to a single number to find out.
bl_ixs=np.array([b1_ixs,b2_ixs,b3_ixs])
bl_ixs=np.sort(bl_ixs,axis=0)
these_triplet_nums=(10**(2*nbits))*bl_ixs[2,:]+ (10**nbits)*bl_ixs[1,:]+bl_ixs[0,:]
# Just add them all and remove the duplicates later.
already_done[bsp_ix:bsp_ix+n_ix3s]=these_triplet_nums
# add to all the arrays
uv_to_bsp_line=np.zeros((n_ix3s,self.nbuv))
diag=np.arange(n_ix3s)
uv_to_bsp_line[diag,b1_ixs]+=1
uv_to_bsp_line[diag,b2_ixs]+=1
uv_to_bsp_line[diag,b3_ixs]+=-1
uv_to_bsp[bsp_ix:bsp_ix+n_ix3s,:]=uv_to_bsp_line
bsp_u[bsp_ix:bsp_ix+n_ix3s,:]=np.transpose(np.array([uv1[:,0],uv2[:,0],uv3[:,0]]))
bsp_v[bsp_ix:bsp_ix+n_ix3s,:]=np.transpose(np.array([uv1[:,1],uv2[:,1],uv3[:,1]]))
bsp_ix+=n_ix3s
# remove the duplicates every n loops
if (ix1 % n) == ((self.nbh-1) % n):
# the (nbh-1 mod n) ensures we do this on the last iteration as well
dummy,unique_ix=np.unique(already_done[0:bsp_ix+n_ix3s],return_index=True)
bsp_ix=len(unique_ix)
already_done[0:bsp_ix]=already_done[unique_ix]
already_done[bsp_ix:]=0
uv_to_bsp[0:bsp_ix]=uv_to_bsp[unique_ix]
bsp_u[0:bsp_ix]=bsp_u[unique_ix]
bsp_v[0:bsp_ix]=bsp_v[unique_ix]
# Only print the status every 5*n iterations
if (ix1 % (5*n)) == ((self.nbh-1) % n):
print 'Done',ix1,'of',self.nbh,'. ',bsp_ix,' bispectra found. Time taken:',np.round(time.time()-tstart,decimals=1),'sec'
print 'Done. Total time taken:',np.round((time.time()-tstart)/60.,decimals=1),'mins'
# Remove the excess parts of each array and attach them to the kpi.
nbsp=bsp_ix
self.already_done=already_done
self.nbsp=bsp_ix
self.uv_to_bsp=uv_to_bsp[0:bsp_ix]
self.bsp_u=bsp_u[0:bsp_ix]
self.bsp_v=bsp_v[0:bsp_ix]
print 'Found',nbsp,'bispectra'
t_start2 = time.time()
tol = 1e-5
try:
if bsp_mat == 'sparse':
print 'Doing sparse svd'
rank = np.linalg.matrix_rank(uv_to_bsp.astype('double'), tol = tol)
print 'Matrix rank:',rank
u, s, vt = svds(uv_to_bsp.astype('double').T, k=rank)
elif bsp_mat == 'full':
print 'Attempting full svd'
u, s, vt = np.linalg.svd(uv_to_bsp.astype('double').T,full_matrices=False)
rank = np.sum(s>tol)
sys.stdout.flush()
self.uv_to_bsp_raw = np.copy(uv_to_bsp)
self.uv_to_bsp = u.T
self.nbsp = rank
self.bsp_s = s
print 'Reduced-rank bispectrum matrix calculated.'
print 'Matrix shape',self.uv_to_bsp.shape
print 'Time taken:',np.round((time.time()-t_start2)/60.,decimals=1),'mins'
if verbose:
print np.log(s)
return s
except:
print 'SVD failed. Using raw matrix.'
self.uv_to_bsp = uv_to_bsp
self.nbsp = nbsp
sys.stdout.flush()