bis_avg.py 文件源码

def pool(biz_dict, vlad_dict, mode):
    if mode == 'train':
        y_dict = read_y()
    y = np.zeros((0, 9))
    x = np.array([])
    x_vlad = np.array([])

    for key, value in sorted(biz_dict.items()):
        avg = np.array(value).sum(axis=0) / len(value)
        vlad = vlad_dict.get(key)
#        vlad = preprocessing.normalize(vlad)
#        print(vlad.shape)
#        feat = np.concatenate([avg, vlad], axis=0)
#        feat = preprocessing.Normalizer().fit_transform(feat)
#        feat = avg
        x = np.vstack((x, avg)) if x.size else avg
        x_vlad = np.vstack((x_vlad, vlad)) if x_vlad.size else vlad

        if mode == 'train':
            y = np.vstack((y, y_dict.get(key)))        
    return (x, x_vlad, y) if mode == 'train' else (x, x_vlad)