2.3. Optimize a dense network with Bayesian optimization#
Authors: Javier Duarte, Thong Nguyen
2.3.1. Loading pandas DataFrames#
Now we load two different NumPy arrays. One corresponding to the VV signal and one corresponding to the background.
import uproot
import numpy as np
import pandas as pd
import h5py
# fix random seed for reproducibility
seed = 7
np.random.seed(seed)
treename = "HZZ4LeptonsAnalysisReduced"
filename = {}
upfile = {}
params = {}
df = {}
filename["VV"] = "data/ntuple_4mu_VV.root"
filename["bkg"] = "data/ntuple_4mu_bkg.root"
VARS = ["f_mass4l", "f_massjj"] # choose which vars to use (2d)
upfile["VV"] = uproot.open(filename["VV"])
upfile["bkg"] = uproot.open(filename["bkg"])
df["bkg"] = upfile["bkg"][treename].arrays(VARS, library="pd")
df["VV"] = upfile["VV"][treename].arrays(VARS, library="pd")
# cut out undefined variables VARS[0] and VARS[1] > -999
df["VV"] = df["VV"][(df["VV"][VARS[0]] > -999) & (df["VV"][VARS[1]] > -999)]
df["bkg"] = df["bkg"][(df["bkg"][VARS[0]] > -999) & (df["bkg"][VARS[1]] > -999)]
# add isSignal variable
df["VV"]["isSignal"] = np.ones(len(df["VV"]))
df["bkg"]["isSignal"] = np.zeros(len(df["bkg"]))
2.3.2. Define the model#
We’ll start with a dense (fully-connected) NN layer. Our model will have a single fully-connected hidden layer with the same number of neurons as input variables. The weights are initialized using a small Gaussian random number. We will switch between linear and tanh activation functions for the hidden layer. The output layer contains a single neuron in order to make predictions. It uses the sigmoid activation function in order to produce a probability output in the range of 0 to 1.
We are using the binary_crossentropy loss function during training, a standard loss function for binary classification problems.
We will optimize the model with the Adam algorithm for stochastic gradient descent and we will collect accuracy metrics while the model is trained.
# baseline keras model
from tensorflow.keras.models import Sequential, Model
from tensorflow.keras.optimizers import SGD
from tensorflow.keras.layers import (
Input,
Activation,
Dense,
Convolution2D,
MaxPooling2D,
Dropout,
Flatten,
)
from tensorflow.keras import optimizers
NDIM = len(VARS)
inputs = Input(shape=(NDIM,), name="input")
outputs = Dense(1, name="output", kernel_initializer="normal", activation="sigmoid")(inputs)
# creae the model
model = Model(inputs=inputs, outputs=outputs)
# compile the model
model.compile(optimizer="adam", loss="binary_crossentropy", metrics=["accuracy"])
# print the model summary
model.summary()
Model: "model"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input (InputLayer) [(None, 2)] 0
_________________________________________________________________
output (Dense) (None, 1) 3
=================================================================
Total params: 3
Trainable params: 3
Non-trainable params: 0
_________________________________________________________________
2022-07-11 21:53:58.557604: I tensorflow/core/platform/cpu_feature_guard.cc:142] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations: SSE4.1 SSE4.2 AVX AVX2 FMA
To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.
2.3.3. Dividing the data into testing and training dataset#
We will split the data into two parts (one for training+validation and one for testing).
We will also apply “standard scaling” preprocessing: http://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.StandardScaler.html i.e. making the mean = 0 and the RMS = 1 for all input variables (based only on the training/validation dataset).
We will also define our early stopping criteria to prevent over-fitting and we will save the model based on the best val_loss.
df_all = pd.concat([df["VV"], df["bkg"]])
dataset = df_all.values
X = dataset[:, 0:NDIM]
Y = dataset[:, NDIM]
from sklearn.model_selection import train_test_split
X_train_val, X_test, Y_train_val, Y_test = train_test_split(X, Y, test_size=0.2, random_state=7)
# preprocessing: standard scalar
from sklearn.preprocessing import StandardScaler
scaler = StandardScaler().fit(X_train_val)
X_train_val = scaler.transform(X_train_val)
X_test = scaler.transform(X_test)
# early stopping callback
from tensorflow.keras.callbacks import EarlyStopping
early_stopping = EarlyStopping(monitor="val_loss", patience=10)
# model checkpoint callback
# this saves our model architecture + parameters into dense_model.h5
from tensorflow.keras.callbacks import ModelCheckpoint
model_checkpoint = ModelCheckpoint(
"dense_model.h5",
monitor="val_loss",
verbose=0,
save_best_only=True,
save_weights_only=False,
mode="auto",
period=1,
)
WARNING:tensorflow:`period` argument is deprecated. Please use `save_freq` to specify the frequency in number of batches seen.
2.3.4. Run training#
Here, we run the training.
# Train classifier
import time
begt = time.time()
history = model.fit(
X_train_val,
Y_train_val,
epochs=100,
batch_size=1024,
verbose=0, # switch to 1 for more verbosity
callbacks=[early_stopping, model_checkpoint],
validation_split=0.25,
)
print("Finished in {}s".format(time.time() - begt))
2022-07-11 21:54:07.493636: I tensorflow/compiler/mlir/mlir_graph_optimization_pass.cc:176] None of the MLIR Optimization Passes are enabled (registered 2)
2022-07-11 21:54:07.494348: I tensorflow/core/platform/profile_utils/cpu_utils.cc:114] CPU Frequency: 2099920000 Hz
Finished in 4.753435134887695s
2.3.5. Optimize the hyperparameters of the model#
The hyperperparameters of the model that we weill optimize are the number of hidden layers num_hidden, the number of nodes in each layer initial_node, and the fraction of dropout dropout.
from skopt import gp_minimize
from skopt.space import Real, Integer
from skopt.utils import use_named_args
def build_custom_model(num_hiddens=2, initial_node=50, dropout=0.5):
inputs = Input(shape=(NDIM,), name="input")
hidden = None
for i in range(num_hiddens):
hidden = Dense(int(round(initial_node / np.power(2, i))), activation="relu")(
inputs if i == 0 else hidden
)
hidden = Dropout(np.float32(dropout))(hidden)
outputs = Dense(1, name="output", kernel_initializer="normal", activation="sigmoid")(hidden)
model = Model(inputs=inputs, outputs=outputs)
return model
def train(batch_size=1000):
history = model.fit(
X_train_val,
Y_train_val,
epochs=100,
batch_size=batch_size,
verbose=0, # switch to 1 for more verbosity
callbacks=[early_stopping, model_checkpoint],
validation_split=0.25,
)
best_acc = max(history.history["val_accuracy"])
return best_acc
space = [
Integer(1, 3, name="hidden_layers"),
Integer(5, 100, name="initial_nodes"),
Real(0.0, 0.9, name="dropout"),
Integer(500, 5000, name="batch_size"),
Real(10**-5, 10**-1, "log-uniform", name="learning_rate"),
]
@use_named_args(space)
def objective(**X):
print("New configuration: {}".format(X))
model = build_custom_model(
num_hiddens=X["hidden_layers"], initial_node=X["initial_nodes"], dropout=X["dropout"]
)
model.compile(
optimizer=optimizers.Adam(lr=X["learning_rate"]),
loss="binary_crossentropy",
metrics=["accuracy"],
)
model.summary()
best_acc = train(batch_size=X["batch_size"])
print("Best acc: {}".format(best_acc))
return -best_acc
begt = time.time()
res_gp = gp_minimize(objective, space, n_calls=20, random_state=3)
print("Finish optimization in {}s".format(time.time() - begt))
New configuration: {'hidden_layers': 1, 'initial_nodes': 85, 'dropout': 0.10919572308469448, 'batch_size': 3062, 'learning_rate': 0.0005600770399424082}
Model: "model_1"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input (InputLayer) [(None, 2)] 0
_________________________________________________________________
dense (Dense) (None, 85) 255
_________________________________________________________________
dropout (Dropout) (None, 85) 0
_________________________________________________________________
output (Dense) (None, 1) 86
=================================================================
Total params: 341
Trainable params: 341
Non-trainable params: 0
_________________________________________________________________
/home/cms.rkansal/miniconda3/envs/machine-learning-hats-2022/lib/python3.9/site-packages/tensorflow/python/keras/optimizer_v2/optimizer_v2.py:374: UserWarning: The `lr` argument is deprecated, use `learning_rate` instead.
warnings.warn(
Best acc: 0.9269527196884155
New configuration: {'hidden_layers': 1, 'initial_nodes': 9, 'dropout': 0.22309946750343207, 'batch_size': 921, 'learning_rate': 0.006015816920825813}
Model: "model_2"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input (InputLayer) [(None, 2)] 0
_________________________________________________________________
dense_1 (Dense) (None, 9) 27
_________________________________________________________________
dropout_1 (Dropout) (None, 9) 0
_________________________________________________________________
output (Dense) (None, 1) 10
=================================================================
Total params: 37
Trainable params: 37
Non-trainable params: 0
_________________________________________________________________
Best acc: 0.9390067458152771
New configuration: {'hidden_layers': 1, 'initial_nodes': 48, 'dropout': 0.19401930757342695, 'batch_size': 2093, 'learning_rate': 0.0009344233883787171}
Model: "model_3"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input (InputLayer) [(None, 2)] 0
_________________________________________________________________
dense_2 (Dense) (None, 48) 144
_________________________________________________________________
dropout_2 (Dropout) (None, 48) 0
_________________________________________________________________
output (Dense) (None, 1) 49
=================================================================
Total params: 193
Trainable params: 193
Non-trainable params: 0
_________________________________________________________________
Best acc: 0.9402121305465698
New configuration: {'hidden_layers': 3, 'initial_nodes': 78, 'dropout': 0.8762834402421655, 'batch_size': 2311, 'learning_rate': 0.001625877427086297}
Model: "model_4"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input (InputLayer) [(None, 2)] 0
_________________________________________________________________
dense_3 (Dense) (None, 78) 234
_________________________________________________________________
dropout_3 (Dropout) (None, 78) 0
_________________________________________________________________
dense_4 (Dense) (None, 39) 3081
_________________________________________________________________
dropout_4 (Dropout) (None, 39) 0
_________________________________________________________________
dense_5 (Dense) (None, 20) 800
_________________________________________________________________
dropout_5 (Dropout) (None, 20) 0
_________________________________________________________________
output (Dense) (None, 1) 21
=================================================================
Total params: 4,136
Trainable params: 4,136
Non-trainable params: 0
_________________________________________________________________
Best acc: 0.9440694451332092
New configuration: {'hidden_layers': 2, 'initial_nodes': 61, 'dropout': 0.2474280963926089, 'batch_size': 2546, 'learning_rate': 0.010731513601673247}
Model: "model_5"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input (InputLayer) [(None, 2)] 0
_________________________________________________________________
dense_6 (Dense) (None, 61) 183
_________________________________________________________________
dropout_6 (Dropout) (None, 61) 0
_________________________________________________________________
dense_7 (Dense) (None, 30) 1860
_________________________________________________________________
dropout_7 (Dropout) (None, 30) 0
_________________________________________________________________
output (Dense) (None, 1) 31
=================================================================
Total params: 2,074
Trainable params: 2,074
Non-trainable params: 0
_________________________________________________________________
Best acc: 0.9457569718360901
New configuration: {'hidden_layers': 3, 'initial_nodes': 29, 'dropout': 0.8037585947601061, 'batch_size': 3733, 'learning_rate': 1.2090882997080298e-05}
Model: "model_6"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input (InputLayer) [(None, 2)] 0
_________________________________________________________________
dense_8 (Dense) (None, 29) 87
_________________________________________________________________
dropout_8 (Dropout) (None, 29) 0
_________________________________________________________________
dense_9 (Dense) (None, 14) 420
_________________________________________________________________
dropout_9 (Dropout) (None, 14) 0
_________________________________________________________________
dense_10 (Dense) (None, 7) 105
_________________________________________________________________
dropout_10 (Dropout) (None, 7) 0
_________________________________________________________________
output (Dense) (None, 1) 8
=================================================================
Total params: 620
Trainable params: 620
Non-trainable params: 0
_________________________________________________________________
Best acc: 0.9469624161720276
New configuration: {'hidden_layers': 1, 'initial_nodes': 36, 'dropout': 0.7280362834123955, 'batch_size': 2295, 'learning_rate': 0.000499811722314133}
Model: "model_7"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input (InputLayer) [(None, 2)] 0
_________________________________________________________________
dense_11 (Dense) (None, 36) 108
_________________________________________________________________
dropout_11 (Dropout) (None, 36) 0
_________________________________________________________________
output (Dense) (None, 1) 37
=================================================================
Total params: 145
Trainable params: 145
Non-trainable params: 0
_________________________________________________________________
Best acc: 0.9491320848464966
New configuration: {'hidden_layers': 2, 'initial_nodes': 80, 'dropout': 0.6422091724381469, 'batch_size': 604, 'learning_rate': 0.00018326055021161485}
Model: "model_8"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input (InputLayer) [(None, 2)] 0
_________________________________________________________________
dense_12 (Dense) (None, 80) 240
_________________________________________________________________
dropout_12 (Dropout) (None, 80) 0
_________________________________________________________________
dense_13 (Dense) (None, 40) 3240
_________________________________________________________________
dropout_13 (Dropout) (None, 40) 0
_________________________________________________________________
output (Dense) (None, 1) 41
=================================================================
Total params: 3,521
Trainable params: 3,521
Non-trainable params: 0
_________________________________________________________________
Best acc: 0.9505785703659058
New configuration: {'hidden_layers': 2, 'initial_nodes': 76, 'dropout': 0.5486612987713032, 'batch_size': 1415, 'learning_rate': 0.004116455326627872}
Model: "model_9"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input (InputLayer) [(None, 2)] 0
_________________________________________________________________
dense_14 (Dense) (None, 76) 228
_________________________________________________________________
dropout_14 (Dropout) (None, 76) 0
_________________________________________________________________
dense_15 (Dense) (None, 38) 2926
_________________________________________________________________
dropout_15 (Dropout) (None, 38) 0
_________________________________________________________________
output (Dense) (None, 1) 39
=================================================================
Total params: 3,193
Trainable params: 3,193
Non-trainable params: 0
_________________________________________________________________
Best acc: 0.9508196711540222
New configuration: {'hidden_layers': 2, 'initial_nodes': 88, 'dropout': 0.730104980246572, 'batch_size': 2184, 'learning_rate': 1.9989431770225823e-05}
Model: "model_10"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input (InputLayer) [(None, 2)] 0
_________________________________________________________________
dense_16 (Dense) (None, 88) 264
_________________________________________________________________
dropout_16 (Dropout) (None, 88) 0
_________________________________________________________________
dense_17 (Dense) (None, 44) 3916
_________________________________________________________________
dropout_17 (Dropout) (None, 44) 0
_________________________________________________________________
output (Dense) (None, 1) 45
=================================================================
Total params: 4,225
Trainable params: 4,225
Non-trainable params: 0
_________________________________________________________________
Best acc: 0.9513018131256104
New configuration: {'hidden_layers': 3, 'initial_nodes': 65, 'dropout': 0.4067110900503197, 'batch_size': 4536, 'learning_rate': 1.6909401832406566e-05}
Model: "model_11"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input (InputLayer) [(None, 2)] 0
_________________________________________________________________
dense_18 (Dense) (None, 65) 195
_________________________________________________________________
dropout_18 (Dropout) (None, 65) 0
_________________________________________________________________
dense_19 (Dense) (None, 32) 2112
_________________________________________________________________
dropout_19 (Dropout) (None, 32) 0
_________________________________________________________________
dense_20 (Dense) (None, 16) 528
_________________________________________________________________
dropout_20 (Dropout) (None, 16) 0
_________________________________________________________________
output (Dense) (None, 1) 17
=================================================================
Total params: 2,852
Trainable params: 2,852
Non-trainable params: 0
_________________________________________________________________
/home/cms.rkansal/miniconda3/envs/machine-learning-hats-2022/lib/python3.9/site-packages/tensorflow/python/keras/optimizer_v2/optimizer_v2.py:374: UserWarning: The `lr` argument is deprecated, use `learning_rate` instead.
warnings.warn(
Best acc: 0.9513018131256104
New configuration: {'hidden_layers': 1, 'initial_nodes': 44, 'dropout': 0.46309479901862954, 'batch_size': 556, 'learning_rate': 0.0798827198804181}
Model: "model_12"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input (InputLayer) [(None, 2)] 0
_________________________________________________________________
dense_21 (Dense) (None, 44) 132
_________________________________________________________________
dropout_21 (Dropout) (None, 44) 0
_________________________________________________________________
output (Dense) (None, 1) 45
=================================================================
Total params: 177
Trainable params: 177
Non-trainable params: 0
_________________________________________________________________
/home/cms.rkansal/miniconda3/envs/machine-learning-hats-2022/lib/python3.9/site-packages/tensorflow/python/keras/optimizer_v2/optimizer_v2.py:374: UserWarning: The `lr` argument is deprecated, use `learning_rate` instead.
warnings.warn(
Best acc: 0.9515429139137268
New configuration: {'hidden_layers': 3, 'initial_nodes': 100, 'dropout': 0.4067987179973009, 'batch_size': 5000, 'learning_rate': 0.1}
Model: "model_13"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input (InputLayer) [(None, 2)] 0
_________________________________________________________________
dense_22 (Dense) (None, 100) 300
_________________________________________________________________
dropout_22 (Dropout) (None, 100) 0
_________________________________________________________________
dense_23 (Dense) (None, 50) 5050
_________________________________________________________________
dropout_23 (Dropout) (None, 50) 0
_________________________________________________________________
dense_24 (Dense) (None, 25) 1275
_________________________________________________________________
dropout_24 (Dropout) (None, 25) 0
_________________________________________________________________
output (Dense) (None, 1) 26
=================================================================
Total params: 6,651
Trainable params: 6,651
Non-trainable params: 0
_________________________________________________________________
/home/cms.rkansal/miniconda3/envs/machine-learning-hats-2022/lib/python3.9/site-packages/tensorflow/python/keras/optimizer_v2/optimizer_v2.py:374: UserWarning: The `lr` argument is deprecated, use `learning_rate` instead.
warnings.warn(
Best acc: 0.9515429139137268
New configuration: {'hidden_layers': 3, 'initial_nodes': 100, 'dropout': 0.4982381104731523, 'batch_size': 5000, 'learning_rate': 1e-05}
Model: "model_14"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input (InputLayer) [(None, 2)] 0
_________________________________________________________________
dense_25 (Dense) (None, 100) 300
_________________________________________________________________
dropout_25 (Dropout) (None, 100) 0
_________________________________________________________________
dense_26 (Dense) (None, 50) 5050
_________________________________________________________________
dropout_26 (Dropout) (None, 50) 0
_________________________________________________________________
dense_27 (Dense) (None, 25) 1275
_________________________________________________________________
dropout_27 (Dropout) (None, 25) 0
_________________________________________________________________
output (Dense) (None, 1) 26
=================================================================
Total params: 6,651
Trainable params: 6,651
Non-trainable params: 0
_________________________________________________________________
/home/cms.rkansal/miniconda3/envs/machine-learning-hats-2022/lib/python3.9/site-packages/tensorflow/python/keras/optimizer_v2/optimizer_v2.py:374: UserWarning: The `lr` argument is deprecated, use `learning_rate` instead.
warnings.warn(
Best acc: 0.9515429139137268
New configuration: {'hidden_layers': 1, 'initial_nodes': 14, 'dropout': 0.34733425561414655, 'batch_size': 601, 'learning_rate': 0.00017779328843182263}
Model: "model_15"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input (InputLayer) [(None, 2)] 0
_________________________________________________________________
dense_28 (Dense) (None, 14) 42
_________________________________________________________________
dropout_28 (Dropout) (None, 14) 0
_________________________________________________________________
output (Dense) (None, 1) 15
=================================================================
Total params: 57
Trainable params: 57
Non-trainable params: 0
_________________________________________________________________
/home/cms.rkansal/miniconda3/envs/machine-learning-hats-2022/lib/python3.9/site-packages/tensorflow/python/keras/optimizer_v2/optimizer_v2.py:374: UserWarning: The `lr` argument is deprecated, use `learning_rate` instead.
warnings.warn(
Best acc: 0.9517840147018433
New configuration: {'hidden_layers': 3, 'initial_nodes': 63, 'dropout': 5.503648636758786e-05, 'batch_size': 2413, 'learning_rate': 2.260306625151787e-05}
Model: "model_16"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input (InputLayer) [(None, 2)] 0
_________________________________________________________________
dense_29 (Dense) (None, 63) 189
_________________________________________________________________
dropout_29 (Dropout) (None, 63) 0
_________________________________________________________________
dense_30 (Dense) (None, 32) 2048
_________________________________________________________________
dropout_30 (Dropout) (None, 32) 0
_________________________________________________________________
dense_31 (Dense) (None, 16) 528
_________________________________________________________________
dropout_31 (Dropout) (None, 16) 0
_________________________________________________________________
output (Dense) (None, 1) 17
=================================================================
Total params: 2,782
Trainable params: 2,782
Non-trainable params: 0
_________________________________________________________________
/home/cms.rkansal/miniconda3/envs/machine-learning-hats-2022/lib/python3.9/site-packages/tensorflow/python/keras/optimizer_v2/optimizer_v2.py:374: UserWarning: The `lr` argument is deprecated, use `learning_rate` instead.
warnings.warn(
Best acc: 0.9517840147018433
New configuration: {'hidden_layers': 1, 'initial_nodes': 93, 'dropout': 0.9, 'batch_size': 4594, 'learning_rate': 0.07356215620398546}
Model: "model_17"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input (InputLayer) [(None, 2)] 0
_________________________________________________________________
dense_32 (Dense) (None, 93) 279
_________________________________________________________________
dropout_32 (Dropout) (None, 93) 0
_________________________________________________________________
output (Dense) (None, 1) 94
=================================================================
Total params: 373
Trainable params: 373
Non-trainable params: 0
_________________________________________________________________
/home/cms.rkansal/miniconda3/envs/machine-learning-hats-2022/lib/python3.9/site-packages/tensorflow/python/keras/optimizer_v2/optimizer_v2.py:374: UserWarning: The `lr` argument is deprecated, use `learning_rate` instead.
warnings.warn(
Best acc: 0.9517840147018433
New configuration: {'hidden_layers': 3, 'initial_nodes': 23, 'dropout': 0.18499533850346483, 'batch_size': 4459, 'learning_rate': 0.0005377686889554435}
Model: "model_18"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input (InputLayer) [(None, 2)] 0
_________________________________________________________________
dense_33 (Dense) (None, 23) 69
_________________________________________________________________
dropout_33 (Dropout) (None, 23) 0
_________________________________________________________________
dense_34 (Dense) (None, 12) 288
_________________________________________________________________
dropout_34 (Dropout) (None, 12) 0
_________________________________________________________________
dense_35 (Dense) (None, 6) 78
_________________________________________________________________
dropout_35 (Dropout) (None, 6) 0
_________________________________________________________________
output (Dense) (None, 1) 7
=================================================================
Total params: 442
Trainable params: 442
Non-trainable params: 0
_________________________________________________________________
/home/cms.rkansal/miniconda3/envs/machine-learning-hats-2022/lib/python3.9/site-packages/tensorflow/python/keras/optimizer_v2/optimizer_v2.py:374: UserWarning: The `lr` argument is deprecated, use `learning_rate` instead.
warnings.warn(
Best acc: 0.9515429139137268
New configuration: {'hidden_layers': 2, 'initial_nodes': 22, 'dropout': 0.8998211088920843, 'batch_size': 639, 'learning_rate': 0.007478381525904193}
Model: "model_19"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input (InputLayer) [(None, 2)] 0
_________________________________________________________________
dense_36 (Dense) (None, 22) 66
_________________________________________________________________
dropout_36 (Dropout) (None, 22) 0
_________________________________________________________________
dense_37 (Dense) (None, 11) 253
_________________________________________________________________
dropout_37 (Dropout) (None, 11) 0
_________________________________________________________________
output (Dense) (None, 1) 12
=================================================================
Total params: 331
Trainable params: 331
Non-trainable params: 0
_________________________________________________________________
/home/cms.rkansal/miniconda3/envs/machine-learning-hats-2022/lib/python3.9/site-packages/tensorflow/python/keras/optimizer_v2/optimizer_v2.py:374: UserWarning: The `lr` argument is deprecated, use `learning_rate` instead.
warnings.warn(
Best acc: 0.9517840147018433
New configuration: {'hidden_layers': 2, 'initial_nodes': 26, 'dropout': 0.42163653305477045, 'batch_size': 4824, 'learning_rate': 1.4616042209619643e-05}
Model: "model_20"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input (InputLayer) [(None, 2)] 0
_________________________________________________________________
dense_38 (Dense) (None, 26) 78
_________________________________________________________________
dropout_38 (Dropout) (None, 26) 0
_________________________________________________________________
dense_39 (Dense) (None, 13) 351
_________________________________________________________________
dropout_39 (Dropout) (None, 13) 0
_________________________________________________________________
output (Dense) (None, 1) 14
=================================================================
Total params: 443
Trainable params: 443
Non-trainable params: 0
_________________________________________________________________
/home/cms.rkansal/miniconda3/envs/machine-learning-hats-2022/lib/python3.9/site-packages/tensorflow/python/keras/optimizer_v2/optimizer_v2.py:374: UserWarning: The `lr` argument is deprecated, use `learning_rate` instead.
warnings.warn(
Best acc: 0.9517840147018433
Finish optimization in 53.88740849494934s
2.3.6. Visualize the improvement#
Let’s see how Bayesian optimization improves the accuracy
from skopt.plots import plot_convergence
plot_convergence(res_gp)
<AxesSubplot:title={'center':'Convergence plot'}, xlabel='Number of calls $n$', ylabel='$\\min f(x)$ after $n$ calls'>
print(
"Best parameters: \
\nbest_hidden_layers = {} \
\nbest_initial_nodes = {} \
\nbest_dropout = {} \
\nbest_batch_size = {} \
\nbest_learning_rate = {}".format(
res_gp.x[0], res_gp.x[1], res_gp.x[2], res_gp.x[3], res_gp.x[4]
)
)
Best parameters:
best_hidden_layers = 1
best_initial_nodes = 14
best_dropout = 0.34733425561414655
best_batch_size = 601
best_learning_rate = 0.00017779328843182263
2.3.7. Plot performance#
Here, we plot the history of the training and the performance in a ROC curve
import matplotlib.pyplot as plt
%matplotlib inline
# plot loss vs epoch
plt.figure(figsize=(15, 10))
ax = plt.subplot(2, 2, 1)
ax.plot(history.history["loss"], label="loss")
ax.plot(history.history["val_loss"], label="val_loss")
ax.legend(loc="upper right")
ax.set_xlabel("epoch")
ax.set_ylabel("loss")
# plot accuracy vs epoch
ax = plt.subplot(2, 2, 2)
ax.plot(history.history["accuracy"], label="acc")
ax.plot(history.history["val_accuracy"], label="val_acc")
ax.legend(loc="upper left")
ax.set_xlabel("epoch")
ax.set_ylabel("accuracy")
# Plot ROC
Y_predict = model.predict(X_test)
from sklearn.metrics import roc_curve, auc
fpr, tpr, thresholds = roc_curve(Y_test, Y_predict)
roc_auc = auc(fpr, tpr)
ax = plt.subplot(2, 2, 3)
ax.plot(fpr, tpr, lw=2, color="cyan", label="auc = %.3f" % (roc_auc))
ax.plot([0, 1], [0, 1], linestyle="--", lw=2, color="k", label="random chance")
ax.set_xlim([0, 1.0])
ax.set_ylim([0, 1.0])
ax.set_xlabel("false positive rate")
ax.set_ylabel("true positive rate")
ax.set_title("receiver operating curve")
ax.legend(loc="lower right")
plt.show()
