# count number of parameters in the model
numParams = model.count_params()
# set a valid path for your system to record model checkpoints
checkpointer = ModelCheckpoint(filepath='/tmp/checkpoint.h5', verbose=1,
save_best_only=True)
# Weighted loss
weight_0 = 1/(len([y for y in y_train_valid if y == 0]))
weight_1 = 1/(len([y for y in y_train_valid if y == 1]))
class_weights = {0: weight_0, 1: weight_1}
#
# # fit the model
fittedModel = model.fit(X_train, y_train, batch_size=34, epochs=100,
verbose=2, validation_data=(X_valid, y_valid),
callbacks=[checkpointer], class_weight=class_weights)
# load optimal weights
model.load_weights('/tmp/checkpoint.h5')
# Evaluate
y_probs = model.predict(X_test)
y_pred = y_probs.argmax(axis=-1)
# save score
csv = pd.read_csv('./data/benchmark.csv')
csv['Prediction'] = y_probs
csv.to_csv('./submission/submissionEEGNET.csv', index=False)
print('--------------------Submission file has been generated.--------------------------')
# fit the model. Due to very small sample sizes this can get
# pretty noisy run-to-run, but most runs should be comparable to xDAWN +
# Riemannian geometry classification (below)
################################################################################
hist = model.fit(X_train,
Y_train,
batch_size=16,
epochs=100,
verbose=2,
validation_split=0.33,
shuffle=True,
callbacks=[checkpointer],
class_weight=class_weights)
# load optimal weights
model.load_weights('/content/gdrive/MyDrive/checkpoint.h5')
###############################################################################
# can alternatively used the weights provided in the repo. If so it should get
# you 93% accuracy. Change the WEIGHTS_PATH variable to wherever it is on your
# system.
###############################################################################
# WEIGHTS_PATH = /path/to/EEGNet-8-2-weights.h5
# model.load_weights(WEIGHTS_PATH)
###############################################################################
# make prediction on test set.
###############################################################################
probs = model.predict(X_test)
# Syntax is {class_1:weight_1, class_2:weight_2,...}.
# Weighted loss
weight_0 = 1/(len([y for y in y_train_valid if y == 0]))
weight_1 = 1/(len([y for y in y_train_valid if y == 1]))
class_weights = {0:weight_0, 1:weight_1}
################################################################################
# Fit the model. Due to very small sample sizes this can get
# pretty noisy run-to-run.
################################################################################
fittedModel = model.fit(X_train, Y_train, batch_size = 34, epochs = 100,
verbose = 2, validation_data=(X_valid, Y_valid),
callbacks=[checkpointer], class_weight = class_weights)
# Load optimal weights
model.load_weights(filepath)
###############################################################################
# Make prediction on test set.
###############################################################################
probs = model.predict(X_test)
preds = probs.argmax(axis = -1)
acc = np.mean(preds == y_test.argmax(axis=-1))
auc = roc_auc_score(y_test,preds)
print("Classification Accuracy: %f " % (acc))
print("Area Under Curve: %f" % (auc))
# the syntax is {class_1:weight_1, class_2:weight_2,...}. Here just setting
# the weights all to be 1
class_weights = {0 :1, 1 :1}
################################################################################
# fit the model. Due to very small sample sizes this can get
# pretty noisy run-to-run, but most runs should be comparable to xDAWN +
# Riemannian geometry classification (below)
################################################################################
hist = model.fit(X_train, Y_train, batch_size = 16, epochs = 150,
verbose = 2, validation_split=0.33, shuffle=True,
callbacks=[checkpointer], class_weight = class_weights)
# load optimal weights
model.load_weights('C:\\Users\\USER\\PycharmProjects\\eegnet\\tmp\\checkpoint2.h5')
###############################################################################
# can alternatively used the weights provided in the repo. If so it should get
# you 93% accuracy. Change the WEIGHTS_PATH variable to wherever it is on your
# system.
###############################################################################
# WEIGHTS_PATH = /path/to/EEGNet-8-2-weights.h5
# model.load_weights(WEIGHTS_PATH)
###############################################################################
# make prediction on test set.
###############################################################################
probs = model.predict(X_test)
verbose=2,
validation_data=(X_validate, Y_validate),
callbacks=[checkpointer],
class_weight=class_weights)
# load optimal weights
# model.load_weights('/tmp/checkpoint.h5')
###############################################################################
# can alternatively used the weights provided in the repo. If so it should get
# you 93% accuracy. Change the WEIGHTS_PATH variable to wherever it is on your
# system.
###############################################################################
WEIGHTS_PATH = "EEGNet-8-2-weights.h5"
model.load_weights(WEIGHTS_PATH)
###############################################################################
# make prediction on test set.
###############################################################################
probs = model.predict(X_test)
preds = probs.argmax(axis=-1)
acc = np.mean(preds == Y_test.argmax(axis=-1))
print("Classification accuracy: %f " % (acc))
############################# PyRiemann Portion ##############################
# code is taken from PyRiemann's ERP sample script, which is decoding in
# the tangent space with a logistic regression
# optimization to balance it out. This data is approximately balanced so we
# don't need to do this, but is shown here for illustration/completeness.
##########################################################################
# the syntax is {class_1:weight_1, class_2:weight_2,...}. Here just setting
# the weights all to be 1
class_weights = {0:1, 1:1}
###########################################################################
#fit the model.
###########################################################################
fittedModel = model.fit(X_train, Y_train, batch_size = 16, epochs = 300,
verbose = 2, validation_data=(X_validate, Y_validate),
callbacks=[checkpointer], class_weight = class_weights)
# load optimal weights
model.load_weights('C:/Users/PUBLIC.DESKTOP-8KLP27O/Desktop/SSSEP/SSSEP_data/tmp/checkpoint.h5')
###############################################################################
# can alternatively used the weights provided in the repo. If so it should get
# you 93% accuracy. Change the WEIGHTS_PATH variable to wherever it is on your
# system.
###############################################################################
# WEIGHTS_PATH = /path/to/EEGNet-8-2-weights.h5
# model.load_weights(WEIGHTS_PATH)
###############################################################################
# make prediction on test set.
###############################################################################
probs = model.predict(X_test)
class_weights = {0: 1, 1: 1}
###########################################################################
#fit the model.
###########################################################################
fittedModel = model.fit(X_train,
Y_train,
batch_size=8,
epochs=20,
verbose=2,
validation_data=(X_validate, Y_validate),
callbacks=[checkpointer],
class_weight=class_weights)
# load optimal weights
model.load_weights(r'C:/Users/oo/Desktop/SSSEP/tmp/checkpoint.h5')
###############################################################################
# can alternatively used the weights provided in the repo. If so it should get
# you 93% accuracy. Change the WEIGHTS_PATH variable to wherever it is on your
# system.
###############################################################################
# WEIGHTS_PATH = /path/to/EEGNet-8-2-weights.h5
# model.load_weights(WEIGHTS_PATH)
###############################################################################
# make prediction on test set.
###############################################################################
probs = model.predict(X_test)
# fit the model. Due to very small sample sizes this can get
# pretty noisy run-to-run, but most runs should be comparable to xDAWN +
# Riemannian geometry classification (below)
################################################################################
hist = model.fit(X_train,
Y_train,
batch_size=16,
epochs=500,
verbose=2,
validation_data=(X_validate, Y_validate),
callbacks=[checkpointer],
class_weight=class_weights)
# load optimal weights
model.load_weights(
'C:/MATLAB/DataPreprocessing/DataAnalysis/eeglab_current/tmp/checkpoint.h5'
)
###############################################################################
# can alternatively used the weights provided in the repo. If so it should get
# you 93% accuracy. Change the WEIGHTS_PATH variable to wherever it is on your
# system.
###############################################################################
# WEIGHTS_PATH = /path/to/EEGNet-8-2-weights.h5
# model.load_weights(WEIGHTS_PATH)
###############################################################################
# make prediction on test set.
###############################################################################
# set a valid path for your system to record model checkpoints
checkpointer = ModelCheckpoint(filepath='./checkpoint.h5',
verbose=1,
save_best_only=True)
# the syntax is {class_1:weight_1, class_2:weight_2,...}. Here just setting
# the weights all to be 1
# unused
class_weights = {0: 1, 1: 1, 2: 1, 3: 1}
with tf.device('/device:GPU:0'):
fittedModel = model.fit(X_train,
Y_train,
batch_size=16,
epochs=150,
verbose=2,
validation_data=(X_validate, Y_validate),
callbacks=[checkpointer])
# load optimal weights
model.load_weights('./checkpoint.h5')
###############################################################################
# make prediction on test set.
###############################################################################
probs = model.predict(X_test)
preds = probs.argmax(axis=-1)
acc = np.mean(preds == Y_test.argmax(axis=-1))
print("Classification accuracy: %f " % (acc))