*_feature_*.csv -- feature csv file
        *_trained_model_*.pkl -- trained model

    dog == 1
    cat == 0
        
    MNISTData
        train.csv -- train data from Kaggle
        test.csv -- test data from Kaggle
        valid.csv -- valid data from Kaggle

        METHOD*_trained_model.pkl -- trained model using METHOD*
        pringle_METHOD*.csv     -- Kaggle submission data using METHOD*

    CIFAR-10
        ##### original image ######
        train/ -- 50000 train pictures of size 32*32 (color) , named:Id.jpg
        test/ -- 100000 test pictures of size 32*32

        ##### label-class transform ######
        trainLabels.csv -- Id,Label all 50000 train Ids.
        trainIdCls.csv -- Id,Cls all 50000 train Ids

        ##### raw pixel-(Id/Class) info ######
        train_feature_pixel_v.csv -- Id,pixel[0][0],pixel[0][1],....,50000
        train_feature_Cls_pixelv.csv -- Cls,pixel[0][0],...., 50000

        ##### features that are used to train/test ######
        CIFAR_train_feature.csv -- Cls,pixel[0][0] ,.. , selected randomly from train_feature_Cls_pixelv.csv , 25000
        CIFAR_valid_feature.csv -- Cls,pixel[0][0] ,.. , selected randomly from train_feature_Cls_pixelv.csv , 25000

        CIFAR_train4_feature.csv -- Cls,pixel[0][0] ,.. , selected randomly from train_feature_Cls_pixelv.csv , 40000
        CIFAR_valid1_feature.csv -- Cls,pixel[0][0] ,.. , selected randomly from train_feature_Cls_pixelv.csv , 10000

        ##### trained model ######
        ##### Project_Method_LearningRate_FeatureProperty_EpoNumber ######
        CIFAR_lenet_0.05_w41_ep5000.np.pkl -- w41: whole dataset and train : valid = 4:1
        CIFAR_lenet_0.13_w41_ep5000.np.pkl -- w41: whole dataset and train : valid = 4:1

        ##### test_output ######
        ##### Id,ClassId ######
        CIFAR_lenet_0.13_w41_ep5000.csv

        ##### Kaggle final upload file ######
        CIFAR_upload_lenet_0.13_w41_ep5000.csv