{"id":5562,"name":"基于机器学习神经网络的手写数字识别","user":343,"user_name":"awaLiny","user_avatar":"343.jpeg?t=1660454714","user_budge":"2","user_bio":"（ฅ＾・ﻌ・＾）ฅ","content_update_at":"2024-05-28T01:25:27.79Z","intro":"**Features**\n\n输入：14＊14＝196 元\n\n层１：16 元\n\n层２：16 元\n\n输出：10 元\n\n（只训练了一点点，不准）\n\n( •_•)\u0026gt;⌐■-■\n\n**Statistics**\n\n一共 3594 参数。\n在测试集中损失 0.4066，正确率 75.0%\n\n**Whats More**\n\n整个网络的构建、运行全部基于原版 Scratch，连数据（500 个手写数字）都是使用 Scratch 平台工具制做的（\n\n以后可能会更换效率更高的方式。\n\n训练采用 [梯度下降法，运行于Python 脚本](https://www.ccw.site/post/47afafbc-1f78-41de-a49d-38ba457dc7a6)\n\n**Extras**\n\n文章：[猫咪都能看懂的神经网络原理](https://www.ccw.site/post/47afafbc-1f78-41de-a49d-38ba457dc7a6)\n致谢：[3Blue1Brown《深度学习之神经网络的结构》《多元微积分的本质》](https://space.bilibili.com/88461692)","manual":"绿旗后手写数字，点击 Submit 得到结果。","plays":372,"thumbs":7,"remixes":4,"favorites":2,"state":1,"is_remix":false,"remix_root":0,"remix_parent":0,"img":"","opensource":3,"control":1,"keymode":1,"is_comment_allowed":false,"use_ccep":false,"weight":875,"extensions":"","created_at":"2024-05-28T01:24:30.674Z","first_publish_at":"2024-05-28T01:25:08Z","publish_at":"2024-05-28T04:13:01Z","updated_at":"2024-05-28T01:24:30.674Z","tags":[],"union_member":[],"remix_tree":[{"id":5696,"name":"手写数字识别 alpha","user":1948,"user_name":"SoivetNingKun","user_avatar":"1948_1747655877.jpeg","user_budge":0,"content_update_at":"2024-06-23T10:12:15.891Z","plays":24,"thumbs":0,"img":"","is_remix":true,"is_union":false},{"id":5562,"name":"基于机器学习神经网络的手写数字识别","user":343,"user_name":"awaLiny","user_avatar":"343.jpeg?t=1660454714","user_budge":0,"content_update_at":"2024-05-28T01:25:27.79Z","plays":372,"thumbs":7,"img":"","is_remix":false,"is_union":false}],"has_thumb":false,"has_favorite":false}