CIFAR-10

CIFAR10_NN_color.py

from keras.models import Sequential
from keras.layers import Dense
from keras.optimizers import Adam
from keras.datasets import cifar10
from keras.utils.np_utils import to_categorical
import numpy as np
import time
import matplotlib.pyplot as plt

start_time = time.time()
# データの読み込み
(x_train, y_train), (x_test, y_test) = cifar10.load_data()
# 小数化
x_train = x_train / 255.0
x_test = x_test / 255.0
# データ数
m_train, m_test = x_train.shape[0], x_test.shape[0]
# ベクトル化
x_train, x_test = x_train.reshape(m_train, -1), x_test.reshape(m_test, -1)
# ノルムで標準化
x_train = x_train / np.linalg.norm(x_train, ord=2, axis=1, keepdims=True)
x_test = x_test / np.linalg.norm(x_test, ord=2, axis=1, keepdims=True)
# yをOneHotVector化
y_train, y_test = to_categorical(y_train), to_categorical(y_test)

# モデル
print(x_train.shape) #(50000, 3072)
model = Sequential()
model.add(Dense(768, activation="relu", input_shape=x_train.shape[1:]))
model.add(Dense(192, activation="relu"))
model.add(Dense(10, activation="softmax"))

# コンパイル
model.compile(optimizer=Adam(), loss="categorical_crossentropy", metrics=["accuracy"])
# フィット
history = model.fit(x_train, y_train, batch_size=64, epochs=15).history
# 経過時間
print("Elapsed[s] : ", time.time() - start_time)
# テスト精度
test_eval = model.evaluate(x_test, y_test)
print("train accuracy :", history["acc"][-1])
print("test accuracy :", test_eval[1])
# 訓練誤差のプロット
plt.plot(range(len(history["loss"])), history["loss"], marker=".")
plt.show()

#1.66GB

# epoch=30の場合
#Elapsed[s] :  1366.77405834198
#10000/10000 [==============================] - 2s 167us/step
#train accuracy : 0.68696
#test accuracy : 0.5358

# epoch=15の場合
#Elapsed[s] :  697.2960453033447
#10000/10000 [==============================] - 2s 153us/step
#train accuracy : 0.57604
#test accuracy : 0.5231

#_________________________________________________________________
#Layer (type)                 Output Shape              Param #
#=================================================================
#dense_1 (Dense)              (None, 768)               2360064
#_________________________________________________________________
#dense_2 (Dense)              (None, 192)               147648
#_________________________________________________________________
#dense_3 (Dense)              (None, 10)                1930
#=================================================================
#Total params: 2,509,642
#Trainable params: 2,509,642
#Non-trainable params: 0
#_________________________________________________________________