def km3nnet(x):
x = tf.reshape(x, shape=[-1, 13, 13, 18, 3])
fc = cnn(x)
output = tf.matmul(fc, weight([nodes["l5"], NUM_CLASSES
])) + bias(NUM_CLASSES)
return output
def km3nnet(x):
""" input: event tensor numpy shape num_minitimeslices, 18, 18, 13, 3
output: label prediction shape 3 (one hot encoded)"""
# loop over mini time slices
nodes = {"l1": 25,
"l2": 25,
"l3": 80,
"l4": 40,
"l5": 20}
weights = {"l1": weight([4, 4, 4, 3, nodes["l1"]]),
"l2": weight([3, 3, 3, nodes["l1"], nodes["l2"]]),
"l3": weight([11025, nodes["l3"]]),
"l4": weight([nodes["l3"], nodes["l4"]])}
biases = {"l1": bias(nodes["l1"]),
"l2": bias(nodes["l2"]),
"l3": bias(nodes["l3"]),
"l4": bias(nodes["l4"])}
conv1 = tf.nn.relu(
conv3d(x, weights["l1"]) + biases["l1"])
conv2 = tf.nn.relu(
conv3d(conv1, weights["l2"]) + biases["l2"])
conv2 = maxpool3d(conv2)
elements = np.prod(conv2._shape_as_list()[1:])
fc = tf.reshape(conv2, [-1, elements])
fc = tf.nn.relu(
tf.matmul(fc, weights["l3"]) + biases["l3"])
fc = tf.nn.relu(
tf.matmul(fc, weights["l4"]) + biases["l4"])
fc = tf.reshape(fc, [50, 1, 40])
c = tf.unstack(fc, num_mini_timeslices, 0)
lstm_layer = tf.contrib.rnn.BasicLSTMCell(nodes["l5"], forget_bias=1.)
outputs, _ = tf.contrib.rnn.static_rnn(lstm_layer, c, dtype=tf.float32)
def km3nnet(x):
""" input: event tensor numpy shape num_minitimeslices, 18, 18, 13, 3
output: label prediction shape 3 (one hot encoded)"""
# loop over mini time slices
mini_timeslices = tf.unstack(x, num_mini_timeslices, 1)
stacked_lstm = tf.contrib.rnn.MultiRNNCell([lstm_cell() for _ in range(2)])
outputs, states = tf.contrib.rnn.static_rnn(stacked_lstm,
mini_timeslices,
dtype=tf.float32)
output = tf.reshape(outputs[-1], [-1, 13 * 13 * 18 * 10])
W = weight([13 * 13 * 18 * 10, 100])
b = bias(100)
z = tf.matmul(output, W) + b
K = weight([100, NUM_CLASSES])
k = bias(NUM_CLASSES)
output = tf.matmul(z, K) + k
return output
def km3nnet(x):
""" input: event tensor numpy shape 400, 18, 18, 13, 3
output: Energy, position, direction prediction in shape 7"""
out_time_bin = []
# loop over 400 time slices
for i in range(x._shape_as_list()[1]):
input_slice = x[:,i,:,:,:,:]
fc = cnn(input_slice)
out_time_bin.append(fc)
c = tf.concat(out_time_bin, 1)
lstm_layer = tf.contrib.rnn.BasicLSTMCell(nodes["l5"], forget_bias=1)
outputs, _ = tf.contrib.rnn.static_rnn(lstm_layer, [c], dtype=tf.float32)
output = tf.matmul( outputs[-1], weight([nodes["l5"], NUM_CLASSES])) + bias(NUM_CLASSES)
return output
def km3nnet(x):
""" input: event tensor numpy shape num_minitimeslices, 18, 18, 13, 3
output: label prediction shape 3 (one hot encoded)"""
# loop over mini time slices
mini_timeslices = tf.unstack(x, num_mini_timeslices, 1)
out_time_bin = []
for ts in mini_timeslices:
out_time_bin.append(cnn(ts))
c = tf.concat(out_time_bin, 1)
c = tf.reshape(c, [-1, num_mini_timeslices, nodes["l4"]])
c = tf.unstack(c, num_mini_timeslices, 1)
lstm_layer = tf.contrib.rnn.BasicLSTMCell(nodes["l5"], forget_bias=1.)
outputs, _ = tf.contrib.rnn.static_rnn(lstm_layer, c, dtype=tf.float32)
output = tf.matmul(outputs[-1], weight([nodes["l5"], NUM_CLASSES])) + bias(NUM_CLASSES)
return output
def km3nnet(x):
""" input: event tensor numpy shape num_minitimeslices, 18, 18, 13, 3
output: label prediction shape 3 (one hot encoded)"""
# loop over mini time slices
mini_timeslices = tf.unstack(x, num_mini_timeslices, 1)
out_time_bin = []
for ts in mini_timeslices:
out_time_bin.append(cnn(ts))
stacked_lstm = tf.contrib.rnn.MultiRNNCell([lstm_cell() for _ in range(2)])
outputs, states = tf.contrib.rnn.static_rnn(stacked_lstm, out_time_bin, dtype=tf.float32)
fout = []
for o in outputs:
output = tf.reshape(o, [-1, 4 * 4 * 5 * 10 ])
W = weight([4 * 4 * 5 * 10, NUM_CLASSES])
b = bias(NUM_CLASSES)
output = tf.matmul(output, W) + b
fout.append(tf.nn.softmax(output))
return output, fout
title = 'reconstruction'
EVT_TYPES = ['nueCC', 'anueCC', 'nueNC', 'anueNC', 'numuCC', 'anumuCC', 'nuK40', 'anuK40']
NUM_CLASSES = 7 # energie, x, y, z, dx, dy, dz
x = tf.placeholder(tf.float32, [None, 400, 13, 13, 18, 3], name="X_placeholder")
y = tf.placeholder(tf.float32, [None, NUM_CLASSES], name="Y_placeholder")
keep_prob = tf.placeholder(tf.float32)
nodes = {"l1": 25,
"l2": 35,
"l3": 80,
"l4": 40,
"l5": 20}
weights = {"l1": weight([4, 4, 4, 3, nodes["l1"]]),
"l2": weight([3, 3, 3, nodes["l1"], nodes["l2"]]),
"l3": weight([15435, nodes["l3"]]),
"l4": weight([nodes["l3"], nodes["l4"]])}
biases = {"l1": bias(nodes["l1"]),
"l2": bias(nodes["l2"]),
"l3": bias(nodes["l3"]),
"l4": bias(nodes["l4"])}
def cnn(input_slice):
""" input: event tensor numpy shape 1, 13, 13, 18, 3"""
conv1 = tf.nn.relu(
conv3d(input_slice, weights["l1"]) + biases["l1"])
conv2 = tf.nn.relu(
from tf_help import conv3d, maxpool3d, weight, bias
title = 'three_classes_sum_tot'
EVT_TYPES = [
'nueCC', 'anueCC', 'nueNC', 'anueNC', 'numuCC', 'anumuCC', 'nuK40',
'anuK40'
]
NUM_CLASSES = 3
x = tf.placeholder(tf.float32, [None, 13, 13, 18, 3], name="X_placeholder")
y = tf.placeholder(tf.float32, [None, NUM_CLASSES], name="Y_placeholder")
nodes = {"l1": 25, "l2": 35, "l3": 80, "l4": 40, "l5": 20}
weights = {
"l1": weight([4, 4, 4, 3, nodes["l1"]]),
"l2": weight([3, 3, 3, nodes["l1"], nodes["l2"]]),
"l3": weight([3, 3, 3, nodes["l2"], nodes["l3"]]),
"l4": weight([7 * 7 * 9 * nodes["l3"], nodes["l4"]]),
"l5": weight([nodes["l4"], nodes["l5"]])
}
biases = {
"l1": bias(nodes["l1"]),
"l2": bias(nodes["l2"]),
"l3": bias(nodes["l3"]),
"l4": bias(nodes["l4"]),
"l5": bias(nodes["l5"])
}
from tf_help import conv3d, maxpool3d, weight, bias
from helper_functions import EVT_TYPES, NUM_CLASSES, PATH, NUM_TRAIN_EVENTS
title = 'test'
num_mini_timeslices = 50
x = tf.placeholder(tf.float32, [None, 13, 13, 18, 3], name="X_placeholder")
y = tf.placeholder(tf.float32, [None, NUM_CLASSES], name="Y_placeholder")
keep_prob = tf.placeholder(tf.float32)
learning_rate = tf.placeholder(tf.float32)
nodes = {"l1": 25, "l2": 25, "l3": 80, "l4": 40, "l5": 20}
weights = {
"l1": weight([4, 4, 4, 3, nodes["l1"]]),
"l2": weight([3, 3, 3, nodes["l1"], nodes["l2"]]),
"l3": weight([11025, nodes["l3"]]),
"l4": weight([nodes["l3"], nodes["l4"]])
}
biases = {
"l1": bias(nodes["l1"]),
"l2": bias(nodes["l2"]),
"l3": bias(nodes["l3"]),
"l4": bias(nodes["l4"])
}
def cnn(mini_timeslice):
""" input: event tensor numpy shape 1, 13, 13, 18, 3"""
title = 'temporal_convLSTM'
num_mini_timeslices = 50
x = tf.placeholder(tf.float32, [None, num_mini_timeslices, 13, 13, 18, 3], name="X_placeholder")
y = tf.placeholder(tf.float32, [None, NUM_CLASSES], name="Y_placeholder")
keep_prob = tf.placeholder(tf.float32)
learning_rate = tf.placeholder(tf.float32)
nodes = {"l1": 25,
"l2": 25,
"l3": 25,
"l4": 1,
"l5": 20}
weights = {"l1": weight([4, 4, 4, 3, nodes["l1"]]),
"l2": weight([3, 3, 3, nodes["l1"], nodes["l2"]]),
"l3": weight([3, 3, 3, nodes["l2"], nodes["l3"]]),
"l4": weight([2, 2, 2, nodes["l3"], nodes["l4"]])}
biases = {"l1": bias(nodes["l1"]),
"l2": bias(nodes["l2"]),
"l3": bias(nodes["l3"]),
"l4": bias(nodes["l4"])}
def cnn(mini_timeslice):
""" input: event tensor numpy shape 1, 13, 13, 18, 3"""
conv1 = tf.nn.relu(
conv3d(mini_timeslice, weights["l1"]) + biases["l1"])
conv1 = tf.contrib.layers.batch_norm(conv1)
fc = tf.reshape(conv2, [-1, elements])
fc = tf.nn.relu(
tf.matmul(fc, weights["l3"]) + biases["l3"])
fc = tf.nn.relu(
tf.matmul(fc, weights["l4"]) + biases["l4"])
fc = tf.reshape(fc, [50, 1, 40])
c = tf.unstack(fc, num_mini_timeslices, 0)
lstm_layer = tf.contrib.rnn.BasicLSTMCell(nodes["l5"], forget_bias=1.)
outputs, _ = tf.contrib.rnn.static_rnn(lstm_layer, c, dtype=tf.float32)
output = tf.matmul(outputs[-1], weight([nodes["l5"], NUM_CLASSES])) + bias(NUM_CLASSES)
return tf.reshape(output, [3])
class toy_gen:
def __init__(self, title):
self.title = title
def __call__(self):
while True:
e, l = make_toy(50)
yield e, l
data_set = tf.data.Dataset.from_generator(
