머신러닝스터디/2016/2016_09_03

• Project
  
  • kaggle digit recignizer
    
  • train 데이터의 0번째 컬럼이 y값(785열)이고 test는 y값이 주어지지 않음(784 열)
    
    • 데이터 전처리 필요
      
    • pandas 라이브러리에 익숙하지 않아 각자 input data 핸들링 하는 방법에 대해 알아봄
      


• 0번째 컬럼을 분리하였으나 학습정확도가 10%대
  
  • 학습이 전혀 되지 않은 것..(0~9중에 찍었을 때 1/10 확률로 정답)
    

import pandas as pd
import keras
from sklearn.cross_validation import train_test_split
from keras.utils.np_utils import to_categorical

train = pd.read_csv("./input/train.csv")
test  = pd.read_csv("./input/test.csv")

y_train = train['label'].as_matrix()
X_train = train.drop('label', axis=1).as_matrix()

X_train, X_test, y_train, y_test = train_test_split(X_train, y_train, test_size=0.30)

model = keras.models.Sequential()

model.add(keras.layers.Dense(64, input_dim=28*28, activation='relu'))
model.add(keras.layers.Dropout(0.5))

model.add(keras.layers.Dense(32, activation='relu'))
model.add(keras.layers.Dropout(0.5))

model.add(keras.layers.Dense(16, activation='relu'))
model.add(keras.layers.Dropout(0.5))

model.add(keras.layers.Dense(10, activation='softmax'))

model.compile(loss='categorical_crossentropy', optimizer='adagrad', metrics=['accuracy'])
model.fit(X_train, to_categorical(y_train, 10), nb_epoch=5, batch_size=600)

score = model.evaluate(X_test, to_categorical(y_test, 10), batch_size=700)

print(score)

print(model.predict(X_test))[0]
print(y_test[0])