#
# infoset = np.append(np.load(datadir + 'DS3_pad/' + 'track_infoset.npy'),
# np.load(datadir + 'DS4_pad/' + 'track_infoset.npy'), axis = 0)
dataname = 'DS3/'
ocdbdir = DEFAULTS.ocdbdir
datadir = DEFAULTS.datadir + dataname
plotdir = DEFAULTS.plotdir
X = np.load(datadir + 'tracklet_dataset.npy') / 1024
infoset = np.load(datadir + 'tracklet_infoset.npy')
print("Loaded: %s \n" % datadir)
(X, infoset), (Xv,
valid_infoset), (Xt,
test_infoset) = DATA.TVT_split_(X, infoset)
T = infoset[:, 0].astype(int)
Tv = valid_infoset[:, 0].astype(int)
Tt = test_infoset[:, 0].astype(int)
I = infoset
Iv = valid_infoset
It = test_infoset
(cs_1, cs_2, d1_1, d1_2) = (8, 16, 256, 128)
dropout = 0.45
stamp = datetime.datetime.now().strftime("%d-%m-%H%M%S") + "_"
mname = "track_V_C_chamber_dropout%.2f" % dropout
tensorboard, csvlogger = LOG.logger_(dataname, stamp, mname)
input = Input(shape=X.shape[1:], name="X-in")
from tensorflow.keras.layers import Input, Dense, Dropout, Flatten, Conv2D, MaxPool2D, concatenate, GaussianNoise
matplotlib.rcParams.update({'font.size': 16})
matplotlib.rcParams['text.usetex'] = True
dataname = 'all_tracks_6_tracklets_even_chamber_calib/'
ocdbdir = DEFAULTS.ocdbdir
datadir = DEFAULTS.datadir + dataname
plotdir = DEFAULTS.plotdir
dataset = np.load(datadir + 'track_dataset.npy').reshape(-1, 17, 24, 1)
infoset = np.repeat(np.load(datadir + 'track_infoset.npy'), 6, axis=0)
print("Loaded: %s \n" % datadir)
columns = DEFAULTS.info_cols_tracklet_ + DEFAULTS.ocdb_cols1 + DEFAULTS.ocdb_cols2
(X, infoset), (Xv, valid_infoset), (Xt, test_infoset) = DATA.TVT_split_(
dataset, infoset)
T = infoset[:, 0].astype(int)
Tv = valid_infoset[:, 0].astype(int)
Tt = test_infoset[:, 0].astype(int)
tracklet_pid_model = custom_models.ComplexConvTrackletPID()
tracklet_pid_model.compile(
optimizer=tf.train.AdamOptimizer(learning_rate=0.001),
loss='binary_crossentropy',
metrics=[ML.pion_con],
)
history = tracklet_pid_model.fit(
X,
T,
from TOOLS import PLOT, METRICS, DATA, DEFAULTS
matplotlib.rcParams.update({'font.size': 16})
matplotlib.rcParams['text.usetex'] = True
from keras.models import load_model
from keras.utils import CustomObjectScope
from keras.initializers import glorot_uniform
from scipy.signal import convolve2d
directory = DEFAULTS.datadir + 'DS2/'
raw_data = np.load(directory + '0_tracks.npy')
raw_info = np.load(directory + '0_info_set.npy')
print("Loaded: %s" % directory)
sumtracks = [np.sum(raw_info[:, 6] == i) for i in range(7)] #Sanity check
dataset, infoset, coord = DATA.process_tracklet_(raw_data, raw_info)
X, y = DATA.shuffle_(dataset / 1024, infoset[:, 0])
plt.imshow(X[0][:, :, 0])
modldir = "saved_models/"
conv_sizes1 = [16]
conv_sizes2 = [64]
dense_sizes1 = [1024]
dense_sizes2 = [256]
for conv_size1 in conv_sizes1:
for conv_size2 in conv_sizes2:
for dense_size1 in dense_sizes1:
for dense_size2 in dense_sizes2:
mname = "conv-%d-%d-filters-dense-%d-%d-nodes-" % (
# X = np.append(np.load(datadir + 'DS3/' + 'track_dataset.npy'),
# np.load(datadir + 'DS4/' + 'track_dataset.npy'), axis = 0)
#
# infoset = np.append(np.load(datadir + 'DS3/' + 'track_infoset.npy'),
# np.load(datadir + 'DS4/' + 'track_infoset.npy'), axis = 0)
dataname = 'DS3/'
ocdbdir = DEFAULTS.ocdbdir
datadir = DEFAULTS.datadir + dataname
plotdir = DEFAULTS.plotdir
X = np.load(datadir + 'track_dataset.npy') / 1024
infoset = np.load(datadir + 'track_infoset.npy')
print("Loaded: %s \n" % datadir)
(X, infoset), (Xv, valid_infoset) = DATA.TV_split_(X, infoset)
T = infoset[:, 0].astype(int)
Tv = valid_infoset[:, 0].astype(int)
I = infoset
Iv = valid_infoset
(cs_1, cs_2, d1_1, d1_2) = (8, 16, 128, 64)
dropout = 0.45
stamp = datetime.datetime.now().strftime("%d-%m-%H%M%S") + "_"
mname = "track_V_U_"
tensorboard, csvlogger = LOG.logger_(dataname, stamp, mname)
input = Input(shape=X.shape[1:], name="X-in")
x = Conv2D(cs_1, [2, 3], activation='relu', padding='same')(input)
x = MaxPool2D([2, 2], 2, padding='valid')(x)
import numpy as np
import pandas as pd
from TOOLS import DEFAULTS, DATA
datadir = DEFAULTS.datadir + 'DS1/'#'000%d/all/'%run_no
ocdbdir = DEFAULTS.ocdbdir
raw_data = np.load(datadir + '0_tracks.npy')
raw_info = np.load(datadir + '0_info_set.npy') # det ()14:20) row (20:26) col (26:32) presence (32:38)
dataset, infoset, coordinates = DATA.process_track_(raw_data, raw_info)
dataset, infoset, ocdbinfo = ocdb_track_(dataset, infoset, coordinates, DEFAULTS.ocdbdir)
info_cols = ["label", "nsigmae", "nsigmap", "PT", "{dE}/{dx}", "Momenta [GeV]", "$\\eta$", "$\\theta$", "$\\phi$","run_no", "event", "V0trackID", "track"]
# def calib_track_(dataset, infoset, )
# gainglob = np.zeros((dataset.shape[0],17,24)) #gain for chambers
# gainpads = np.zeros((dataset.shape[0],17,24)) #gain for individual pads
# ocdbinfo = np.zeros((dataset.shape[0], chamber.values.shape[1]))
from tensorflow.keras.models import Model
from tensorflow.keras.layers import Input, Dense, Dropout, Flatten, Conv2D, MaxPool2D, concatenate, GaussianNoise
matplotlib.rcParams.update({'font.size': 16})
matplotlib.rcParams['text.usetex'] = True
dataname = 'DS3/'
ocdbdir = DEFAULTS.ocdbdir
datadir = DEFAULTS.datadir + dataname
plotdir = DEFAULTS.plotdir
X = np.load(datadir + 'tracklet_dataset.npy') / 1024
infoset = np.load(datadir + 'tracklet_infoset.npy')
print("Loaded: %s \n" % datadir)
(X, infoset), (Xv, valid_infoset) = DATA.TV_split_(X, infoset)
T = infoset[:, 0].astype(int)
Tv = valid_infoset[:, 0].astype(int)
I = infoset
Iv = valid_infoset
(cs_1, cs_2, d1_1, d1_2) = (8, 16, 128, 64)
mname = "tracklet_V_U"
stamp = datetime.datetime.now().strftime("%d-%m-%H%M%S") + "_"
tensorboard, csvlogger = LOG.logger_(dataname, stamp, mname)
input = Input(shape=X.shape[1:], ) # name="X-in")
x = Conv2D(cs_1, [2, 3], activation='relu', padding='same')(input)
x = MaxPool2D([2, 2], 2, padding='valid')(x)
x = Conv2D(cs_2, [2, 3], activation='relu', padding='same')(x)
x = MaxPool2D([2, 2], 2, padding='valid')(x)
import numpy as np
import matplotlib.pyplot as plt
from TOOLS import DATA, MODELS, LOG, METRICS
import random, matplotlib, datetime, os
run_no = '000265378/'
dataname = 'test/'
directory = 'data/output/' + run_no + dataname
raw_data = np.load(directory + '0_tracks.npy')
raw_info = np.load(directory + '0_info_set.npy')
print("Loaded: %s" % directory)
sumtracks = [np.sum(raw_info[:, 6] == i) for i in range(7)] #Sanity check
dataset, infoset = DATA.process_1(raw_data, raw_info)
X, y = DATA.shuffle_(dataset / 1024, infoset[:, 0])
print("Electron occurence: %.2f" % (100 * sum(y) / len(y)))
X_[str(l)] = np.append(dataset[:, 0][mask].reshape(-1, l),
dataset[:, 6][mask].reshape(-1, l),
axis=1)
P1_ = {}
P2_ = {}
Xt_ = {}
It_ = {}
epoch = [100] * 6
batch = [2**10] * 6
pioncon1 = np.zeros(6)
Upioncon1 = np.zeros(6)
pioncon2 = np.zeros(6)
Upioncon2 = np.zeros(6)
for l in range(1, 7):
X_[str(l)], I_[str(l)] = DATA.shuffle_(X_[str(l)], I_[str(l)])
(X, I), (Xv, Iv), (Xt, It) = DATA.TVT_split_(X_[str(l)],
I_[str(l)],
test_split=0.35,
valid_split=0.15)
Xt_[str(l)] = Xt
It_[str(l)] = It
T = I[:, 0].astype(int)
Tv = Iv[:, 0].astype(int)
Tt = It[:, 0].astype(int)
print(X.shape)
input = Input(shape=X[:, :l].shape[1:])
x = Dense(36, activation='relu')(input)
x = Dropout(0.5)(x)
x = Dense(216, activation='relu')(x)
from sklearn import preprocessing
from tensorflow.keras import backend as K
from tensorflow.keras.models import Model
from tensorflow.keras.layers import Input, Dense, Dropout, Flatten, Conv2D, MaxPool2D, concatenate, GaussianNoise
matplotlib.rcParams.update({'font.size': 16})
matplotlib.rcParams['text.usetex'] = True
infoset = np.load(DEFAULTS.datadir + 'DS4/' + 'tracklet_infoset.npy')
dataset = np.load(DEFAULTS.modldir + 'P-DS4-tracklet_V_U_wbce5.npy')
(I[:,13] ==5.0).sum()
mask = I[:,13] == 6.0
I = I[mask]
X = X[mask]
(X, I), (Xv, Iv), (Xt, It)= DATA.TVT_split_(X, I)
T = I[:,0].astype(int)
Tv = Iv[:,0].astype(int)
Tt = It[:,0].astype(int)
A = I[:,14:]
Av = Iv[:,14:]
At = It[:,14:]
input_X = Input(shape=X.shape[1:], name="info")
x = Dense(64)(input_X)
x = Dense(8)(x)
output_aux = Dense(1, activation='sigmoid')(x)
model = Model(inputs=input_X, outputs=output_aux)
#model.summary()
model.compile(optimizer=tf.train.AdamOptimizer(learning_rate=1e-3), loss=ML.WBCE,metrics=[ML.pion_con])
# dataname = 'all_tracks_6_tracklets_even_chamber_calib/'
# dataname = 'DS1/'
for i in range(1, 5):
dataname = 'DS%d/' % i
ocdbdir = DEFAULTS.ocdbdir
datadir = DEFAULTS.datadir + dataname
plotdir = DEFAULTS.plotdir
dataset = np.load(datadir + '0_tracks.npy')
infoset = np.load(datadir + '0_info_set.npy')
# dataset, infoset = DATA.shuffle_(dataset, infoset)
print("Loaded: %s \n" % datadir)
dataset, infoset = DATA.process_tracklet_(dataset, infoset)
# dataset, infoset = DATA.ocdb_tracklet_(dataset, infoset, ocdbdir)
infoset = DATA.ocdb_expand_(infoset, ocdbdir)
np.save(datadir + 'tracklet_dataset.npy', dataset)
np.save(datadir + 'tracklet_infoset.npy', infoset)
print(dataset.shape)
# for i in range(1,5):
# dataname = 'DS%d_pad/'%i
# ocdbdir = DEFAULTS.ocdbdir
# datadir = DEFAULTS.datadir + dataname
# plotdir = DEFAULTS.plotdir
#
# dataset = np.load(datadir + '0_tracks.npy')
# infoset = np.load(datadir + '0_info_set.npy')
from tensorflow.keras.models import Model
from tensorflow.keras.layers import Input, Dense, Dropout, Flatten, Conv2D, MaxPool2D, concatenate
import tensorflow as tf
import pandas as pd
from sklearn.decomposition import PCA
from sklearn import preprocessing
from tensorflow.keras.callbacks import TensorBoard, Callback
run_no = '000265378/'
dataname = 'all/'
directory = 'data/output/' + run_no + dataname
raw_data = np.load(directory + '0_tracks.npy')
raw_info = np.load(directory + '0_info_set.npy')
print('Loaded: %s' % directory)
dataset, infoset = DATA.process_1(raw_data, raw_info)
dataset, infoset = DATA.shuffle_(dataset, infoset)
def WBCE(y_true, y_pred, weight=7.0, from_logits=False):
y_pred = tf.cast(y_pred, dtype='float32')
y_true = tf.cast(y_true, y_pred.dtype)
return K.mean(ML.weighted_binary_crossentropy(y_true,
y_pred,
weight=weight,
from_logits=from_logits),
axis=-1)
input = Input(shape=X.shape[1:], ) # name="X-in")
x = Conv2D(cs_1, [2, 3], activation='relu', padding='same')(input)
from tensorflow.keras import backend as K
from tensorflow.keras.models import Model
from tensorflow.keras.layers import Input, Dense, Dropout, Flatten, Conv2D, MaxPool2D, concatenate, GaussianNoise
matplotlib.rcParams.update({'font.size': 16})
matplotlib.rcParams['text.usetex'] = True
dataname = 'DS2/'
ocdbdir = DEFAULTS.ocdbdir
datadir = DEFAULTS.datadir + dataname
plotdir = DEFAULTS.plotdir
raw_data = np.load(datadir + '0_tracks.npy')
raw_info = np.load(datadir + '0_info_set.npy')
print("Loaded: %s \n" % datadir)
dataset, infoset = DATA.process_tracklet_(raw_data, raw_info)
(X, infoset), (Xv, valid_infoset), (Xt, test_infoset) = DATA.TVT_split_(
dataset / 1024, infoset)
T = infoset[:, 0].astype(int)
Tv = valid_infoset[:, 0].astype(int)
Tt = test_infoset[:, 0].astype(int)
I = infoset
Iv = valid_infoset
It = test_infoset # I = infoset[:,nx:ny]
(cs_1, cs_2, d1_1, d1_2) = (8, 16, 128, 64)
weights = [1 / 10, 1.0, 10.0]
stamp = datetime.datetime.now().strftime("%d-%m-%H%M%S") + "_"
thresholds = np.linspace(0, 1, 1000)
import pandas as pd
import seaborn as sns
import random, matplotlib
from TOOLS import DATA, PLOT, DEFAULTS
matplotlib.rcParams.update({'font.size': 16})
matplotlib.rcParams['text.usetex'] = True
ocdbdir = DEFAULTS.ocdbdir
datadir = DEFAULTS.datadir + 'DS2/' #'000%d/all/'%run_no
plotfir = DEFAULTS.plotdir
dataset = np.load(datadir + '0_tracks.npy')
infoset = np.load(datadir + '0_info_set.npy')
print("Loaded: %s \n" % datadir)
dataset, infoset, coordinates = DATA.process_tracklet_(dataset, infoset)
dataset, infoset = DATA.ocdb_tracklet_(dataset, infoset, coordinates,
DEFAULTS.ocdbdir)
ocdb_cols = ["Anode Voltage", "Drift Voltage", "Drift Velocity", "ExB"]
info_cols = [
"label", "nsigmae", "nsigmap", "PT", "{dE}/{dx}", "Momenta [GeV]",
"$\\eta$", "$\\theta$", "$\\phi$", "run_no", "event", "V0trackID", "track"
]
info_cols.extend(ocdb_cols)
infoframe = pd.DataFrame(data=infoset, columns=info_cols)
infoframe.head(20)
elec_data, elec_info = DATA.elec_strip_(dataset, infoset)
pion_data, pion_info = DATA.pion_strip_(dataset, infoset)
for l in range(1,7):
mask = infoset[:,13] == l
I_[str(l)] = infoset[mask][::l,:14]
X_[str(l)] = dataset[:,0][mask].reshape(-1,l)
P_ = {}
Xt_ = {}
It_ = {}
epoch = [100]*6
batch = [2**10]*6
pioncon = np.zeros(6)
Upioncon = np.zeros(6)
for l in range(1,7):
# X_[str(l)], I_[str(l)] = DATA.shuffle_(X_[str(l)], I_[str(l)])
(X, I), (Xv, Iv), (Xt, It) = DATA.TVT_split_(X_[str(l)], I_[str(l)], test_split=0.35, valid_split=0.15)
T = I[:,0].astype(int)
Tv = Iv[:,0].astype(int)
Tt = It[:,0].astype(int)
print(X.shape)
input = Input(shape=X.shape[1:], name="info")
x = Dense(36, activation = 'relu')(input)
x = Dropout(0.5)(x)
x = Dense(216, activation = 'relu')(x)
x = Dropout(0.5)(x)
x = Dense(36, activation = 'relu')(x)
output = Dense(1, activation='sigmoid')(x)
model = Model(inputs=input, outputs=output)
model.compile(optimizer=tf.train.AdamOptimizer(learning_rate=1e-3), loss=ML.WBCE, metrics=['accuracy'])
import numpy as np
import matplotlib.pyplot as plt
from TOOLS import DATA, MODELS, LOG, METRICS
import random, matplotlib, datetime, os
from sklearn.decomposition import PCA
from sklearn import preprocessing
import pandas as pd
run_no = '000265378/'
dataname = 'even/'
directory = 'data/output/' + run_no + dataname
raw_data = np.load(directory + '0_tracks.npy')
raw_info = np.load(directory + '0_info_set.npy')
print("Loaded: %s" % directory)
raw_data, raw_info = DATA.shuffle_(raw_data, raw_info)
columns = ["label", "nsigmae", "nsigmap", "PT", "${dE}/{dx}$", "Momenta [GeV]", "eta", "theta", "phi", "event", "V0trackID", "track"]
cnames = ["$\\pi$","$e$"]
colour = ['r', 'g']
styles = ['--','-.']
T = raw_info[:,0]
nx = 3
ny = 9
columns[nx:ny+1]
params = raw_info[:,nx:ny+1]
parammean = params.sum(axis=0)/params.shape[0]
paramstdv = np.sqrt(np.square((params - parammean)).sum(axis=0)/(params.shape[0]-1))
I = (params - parammean)