In [1]:
import keras
from keras.models import Sequential
from keras.layers import Dense, Dropout, Activation
from keras.optimizers import SGD, Adam
from keras.layers.normalization import BatchNormalization
from read_dataset import read_ceps_with_train_test
import numpy as np
from create_data_ceps import createCepsSS
from read_saved_models import loadCepsStanderdScaler
from matplotlib import pyplot as plt
%matplotlib inline
In [2]:
X_train, X_test, y_train, y_test = read_ceps_with_train_test(recreate_data=True)
In [3]:
def getStanderizedData(data):
data_shape = data.shape
n = data_shape[0]
reshaped_data = data.reshape(n, -1)
saved_ss = loadCepsStanderdScaler()
trasformed_data = saved_ss.transform(reshaped_data)
ret_data = trasformed_data.reshape(data_shape)
return ret_data
In [4]:
X_ss_train = getStanderizedData(X_train)
X_ss_test = getStanderizedData(X_test)
In [5]:
layer_1_outdim = 1000
layer_1_dropout = 0.5
layer_2_outdim = 800
layer_2_dropout = 0.5
layer_3_outdim = 100
layer_3_dropout = 0.3
In [6]:
model = Sequential()
# this layerf from https://keras.io/getting-started/sequntial-model-guide/
# Dense(64) is fully-connected layer with 64 hidden units
model.add(Dense(layer_1_outdim, activation='relu', input_dim=20))
model.add(Dropout(layer_1_dropout))
model.add(BatchNormalization())
model.add(Dense(layer_2_outdim, activation='relu'))
model.add(Dropout(layer_2_dropout))
model.add(BatchNormalization())
model.add(Dense(layer_3_outdim, activation='relu'))
model.add(Dropout(layer_3_dropout))
model.add(BatchNormalization())
model.add(Dense(10, activation='softmax'))
# sgd = SGD(lr=0.001, decay=1e-6, momentum=0.9, nesterov=True)
adam = Adam(lr=0.01, decay=1e-9)
model.compile(loss='categorical_crossentropy',
optimizer=adam,
metrics=['accuracy'])
epoch = 20
In [7]:
result = model.fit(X_ss_train, y_train, epochs=epoch, batch_size=100, validation_split=0.2)
In [8]:
x = range(epoch)
plt.plot(x, result.history['acc'], label='train accuracy')
plt.plot(x, result.history['val_acc'], label='test accuracy')
plt.legend(loc='center left', bbox_to_anchor=(1, 0.5))
score = model.evaluate(X_ss_test, y_test, batch_size=100)
print("")
print(model.metrics_names)
print(score)
In [9]:
model_filepath = "./savedModels/ceps_standerized_dense_model.h5"
model.save(model_filepath)
In [10]:
# adjast param results
"""
note
---
layer_1_outdim = 100
layer_1_dropout = 0.5
layer_2_outdim = 100
layer_2_dropout = 0.4
layer_3_outdim = 50
layer_3_dropout = 0.5
loss: 1.3307 - acc: 0.5479 - val_loss: 1.4328 - val_acc: 0.4250
test['loss', 'acc'][1.3474427759647369, 0.51999998837709427]
---
"""
Out[10]:
In [11]:
file_path = "../data/songData/genres/disco/disco.00006.ceps.npy"
ceps = np.load(file_path)
t_ceps = ceps.transpose()
t_num_ceps = len(t_ceps)
t_ceps_mean = np.mean(t_ceps[int(t_num_ceps / 10):int(t_num_ceps * 9 / 10)], axis=0)
predict_data = np.array([t_ceps_mean])
predict_data = getStanderizedData(predict_data)
In [12]:
predict_result = model.predict(predict_data)
print(predict_result)
print(np.argmax(predict_result))