看了一个大佬的解释,卷积神经网路初探——LeNet-5的原理与手写数字识别的实现,似懂非懂,看文章下的提问感觉也是没弄清楚,求大佬们解释一波,给大佬递茶了


因为LeNet模型希望能够识别ASCII列印字元,而标准ASCII字元列印大小是7*12=84。

..., they were instead designed to represent a stylized image of the corresponding character class drawn on a 7*12 bitmap(hence the number 84).

Such a representation is not particularly useful for recognizing isolated digits, but it is quite useful for recognizing strings of characters taken from the full printable ASCII set.

由于ASCII字元标准的列印字元,是用7*12大小的点阵图表示的。

LeNet在输出层之前最后一个全连接层的大小设置为84,个人以为,是希望每一维特征分别体现标准7*12大小点阵图上每一个像素点的特性。

参考链接:http://yann.lecun.com/exdb/publis/pdf/lecun-01a.pdf


mark 一下,到实验室再看
推荐阅读:
查看原文 >>
相关文章