mirror/collision-analyisis-and-interpolation
1
0
Fork 0
mirror of https://github.com/Findus23/collision-analyisis-and-interpolation.git synced 2024-09-19 15:13:50 +02:00
Code Releases Activity

strongly improve the neural network

Browse source
This commit is contained in:
Lukas Winkler 2019-07-29 13:59:10 +02:00
parent 9f822df29d
commit e155d9ef96
Signed by: lukas
GPG key ID: 54DE4D798D244853
1 changed files with 24 additions and 16 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

40
kerastest.py
View file

@ -8,22 +8,24 @@ from keras.engine.saving import load_model
from keras.layers import Dense
from keras.utils import plot_model
from matplotlib import pyplot as plt
from sklearn.preprocessing import StandardScaler
from CustomScaler import CustomScaler
from simulation_list import SimulationList
simulations = SimulationList.jsonlines_load()
random.shuffle(simulations.simlist)
train_data = [s for s in simulations.simlist]
test_data = [s for s in simulations.simlist if s.type != "original"]
train_data = set([s for s in simulations.simlist])
new_data = [s for s in simulations.simlist if s.type != "original"]
random.seed(1)
test_data = set(random.sample(new_data, int(len(new_data) * 0.2)))
train_data -= test_data
print(len(train_data), len(test_data))
X = np.array(
[[s.alpha, s.v, s.projectile_mass, s.gamma, s.target_water_fraction, s.projectile_water_fraction] for s in
train_data])
scaler = StandardScaler()
scaler = CustomScaler()
scaler.fit(X)
x = scaler.transform(X)
x = scaler.transform_data(X)
print(x.shape)
Y = np.array([s.water_retention_both for s in train_data])
print(Y.shape)
@ -31,30 +33,36 @@ X_test = np.array(
[[s.alpha, s.v, s.projectile_mass, s.gamma, s.target_water_fraction, s.projectile_water_fraction] for s in
test_data])
Y_test = np.array([s.mass_retention_both for s in test_data])
x_test = scaler.transform(X_test)
x_test = scaler.transform_data(X_test)
# print(X_test)
# print(X[0])
# exit()
tbCallBack = keras.callbacks.TensorBoard(log_dir='./logs', histogram_freq=0, batch_size=32, write_graph=True,
write_grads=False, write_images=False, embeddings_freq=0,
random.seed()
tbCallBack = keras.callbacks.TensorBoard(log_dir='./logs/{}'.format(random.randint(0, 100)), histogram_freq=1,
batch_size=32, write_graph=True,
write_grads=True, write_images=True, embeddings_freq=0,
embeddings_layer_names=None, embeddings_metadata=None, embeddings_data=None,
update_freq='epoch')
if os.path.exists("model.hd5") and False:
if os.path.exists("model.hd5"):
model = load_model("model.hd5")
else:
model = Sequential()
model.add(Dense(6, input_dim=6, kernel_initializer='normal', activation='relu'))
model.add(Dense(6, input_dim=6, activation='relu'))
model.add(Dense(4, kernel_initializer='normal', activation='relu'))
model.add(Dense(1, kernel_initializer='normal'))
model.add(Dense(3, kernel_initializer='normal', activation='relu'))
model.add(Dense(1, kernel_initializer='normal', activation="sigmoid"))
model.compile(loss='mean_squared_error', optimizer='adam')
model.summary()
plot_model(model, "model.png", show_shapes=True, show_layer_names=True)
model.fit(x, Y, epochs=400, callbacks=[tbCallBack], validation_split=0.02)
model.fit(x, Y, epochs=200, callbacks=[tbCallBack], validation_data=(x_test, Y_test))
loss = model.evaluate(x_test, Y_test)
print(loss)
if loss > 0.04:
# exit()
...
# print("-------------------------------------")
# exit()
model.save("model.hd5")
@ -62,14 +70,14 @@ else:
xrange = np.linspace(-0.5, 60.5, 100)
yrange = np.linspace(0.5, 5.5, 100)
xgrid, ygrid = np.meshgrid(xrange, yrange)
mcode = 23e10
mcode = 1e23
wpcode = 15 / 100
wtcode = 15 / 100
gammacode = 0.7
testinput = np.array([[np.nan, np.nan, mcode, gammacode, wtcode, wpcode]] * 100 * 100)
testinput[::, 0] = xgrid.flatten()
testinput[::, 1] = ygrid.flatten()
testinput = scaler.transform(testinput)
testinput = scaler.transform_data(testinput)
print(testinput)
print(testinput.shape)