Digit Recognizer

Digit Recognizer，最终digi_vgg16.py在Kaggle得分0.99071，排名23.7%

经典机器学习算法

• DigitRecognizer_scikit.ipynb: 使用SVM（scikit-learn），sharpen数据，经过测试，以32为临界值，可以得到最佳的预测结果

卷积神经网络

VGG16

VGG16模型

• digi_vgg16.py: 最佳结果，云GPU服务器计算得到结果
• digi_vgg16_dropout.py: 尝试加入一个Dropout层，不过结果变差

自定义神经网络

• DigitRecognizer_tensorflow.ipynb: Low & High API
• DigitRecognizer_keras.ipynb，digi_vgg16.py: 2个是一样的，工具不同而已，summary如下

_________________________________________________________________
Layer (type)                 Output Shape              Param #
=================================================================
conv2d_1 (Conv2D)            (None, 28, 28, 32)        832
_________________________________________________________________
conv2d_2 (Conv2D)            (None, 28, 28, 64)        51264
_________________________________________________________________
max_pooling2d_1 (MaxPooling2 (None, 28, 28, 64)        0
_________________________________________________________________
conv2d_3 (Conv2D)            (None, 28, 28, 128)       204928
_________________________________________________________________
conv2d_4 (Conv2D)            (None, 28, 28, 256)       819456
_________________________________________________________________
max_pooling2d_2 (MaxPooling2 (None, 28, 28, 256)       0
_________________________________________________________________
flatten_1 (Flatten)          (None, 200704)            0
_________________________________________________________________
dropout_1 (Dropout)          (None, 200704)            0
_________________________________________________________________
dense_1 (Dense)              (None, 128)               25690240
_________________________________________________________________
dense_2 (Dense)              (None, 10)                1290
=================================================================
Total params: 26,768,010
Trainable params: 26,768,010
Non-trainable params: 0
_________________________________________________________________