<?xml version="1.0" encoding="utf-8" ?><rss version="2.0"><channel><title><![CDATA[HaoChen的博客]]></title><description><![CDATA[let's learn deep learning together]]></description><link>https://blog.youkuaiyun.com/weixin_57028107</link><language>zh-cn</language><generator>https://blog.youkuaiyun.com/</generator><copyright><![CDATA[Copyright &copy; weixin_57028107]]></copyright><item><title><![CDATA[langchian delete 删除大bug：无法删除，删除id后乱重排]]></title><link>https://blog.youkuaiyun.com/weixin_57028107/article/details/132214687</link><guid>https://blog.youkuaiyun.com/weixin_57028107/article/details/132214687</guid><author>weixin_57028107</author><pubDate>Thu, 10 Aug 2023 20:27:08 +0800</pubDate><description><![CDATA[langchain 无法删除delete 大bug 终极解决方案]]></description><category></category></item><item><title><![CDATA[最简单条件生成对抗网络学习(CGAN)]]></title><link>https://blog.youkuaiyun.com/weixin_57028107/article/details/131978901</link><guid>https://blog.youkuaiyun.com/weixin_57028107/article/details/131978901</guid><author>weixin_57028107</author><pubDate>Fri, 28 Jul 2023 22:18:52 +0800</pubDate><description><![CDATA[我们都知道GAN可以生成很逼真的人脸，它是随机生成人脸呀，如果我指定要生成一张女生或男生的脸怎么办，如果我想指定生成一张迪丽热巴的脸怎么办，是不是它就不行了。再比如它可以生成手势图，比如剪刀手，里面有阴影的，有真人的，那如果我们指定要生成阴影的，它也做不到。因为最初的GAN只是以假乱真，还没有监督信号，所以做不到有条件的生成。]]></description><category></category></item><item><title><![CDATA[生成对抗网络核心训练思路-以DCGAN为例]]></title><link>https://blog.youkuaiyun.com/weixin_57028107/article/details/131855142</link><guid>https://blog.youkuaiyun.com/weixin_57028107/article/details/131855142</guid><author>weixin_57028107</author><pubDate>Fri, 21 Jul 2023 23:17:09 +0800</pubDate><description><![CDATA['''定义生成器 Generator。]]></description><category></category></item><item><title><![CDATA[生成对抗网络GAN 最简单入门学习]]></title><link>https://blog.youkuaiyun.com/weixin_57028107/article/details/131720890</link><guid>https://blog.youkuaiyun.com/weixin_57028107/article/details/131720890</guid><author>weixin_57028107</author><pubDate>Fri, 14 Jul 2023 13:06:49 +0800</pubDate><description><![CDATA[最简单GAN入门。最简单对抗生成网络入门]]></description><category></category></item><item><title><![CDATA[手动搭建seq2seq文本翻译模型-365每周深度学习]]></title><link>https://blog.youkuaiyun.com/weixin_57028107/article/details/131229340</link><guid>https://blog.youkuaiyun.com/weixin_57028107/article/details/131229340</guid><author>weixin_57028107</author><pubDate>Thu, 15 Jun 2023 16:25:22 +0800</pubDate><description><![CDATA[手动搭建seq2seq翻译模型并进行实践]]></description><category></category></item><item><title><![CDATA[DenseNet学习]]></title><link>https://blog.youkuaiyun.com/weixin_57028107/article/details/131112560</link><guid>https://blog.youkuaiyun.com/weixin_57028107/article/details/131112560</guid><author>weixin_57028107</author><pubDate>Thu, 08 Jun 2023 20:25:19 +0800</pubDate><description><![CDATA[我们都知道resnet 的方法是把上一层与这层的加起来。那能不能更彻底一点。直接把之前所有的层都加起来。就这样一篇顶刊出现了。DenseNet。]]></description><category></category></item><item><title><![CDATA[残差网络ResNet50学习与训练多分类任务-365每周深度学习J1]]></title><link>https://blog.youkuaiyun.com/weixin_57028107/article/details/130990037</link><guid>https://blog.youkuaiyun.com/weixin_57028107/article/details/130990037</guid><author>weixin_57028107</author><pubDate>Thu, 01 Jun 2023 17:05:36 +0800</pubDate><description><![CDATA[则在20后的梯度几乎都会小于0.1，最终当层数很大时就几乎等于零。而残杀网络则解决了网络不能深下去的问题。因为它在每一层计算梯度的都开了一个洞，它既可以通过这层，也可以不用这层。所以当这层对梯度消失影响很大时，就可以不用。残差网络的作用为解决梯度消失问题。如何通过激活函数后的梯度始终小于1，不妨设上限为0.9，则经过N层后的梯度会小于。点击头像并查看免费下载。]]></description><category></category></item><item><title><![CDATA[torchtext安装报错终极解决方案 and N4:调用与搭建word2vec网络进行训练]]></title><link>https://blog.youkuaiyun.com/weixin_57028107/article/details/130760826</link><guid>https://blog.youkuaiyun.com/weixin_57028107/article/details/130760826</guid><author>weixin_57028107</author><pubDate>Fri, 19 May 2023 14:28:42 +0800</pubDate><description><![CDATA[使用gensim实例化word2vec网络并添加一个全连接层进行文本多标签分类。并学习如何正确安装torchtext在指定gpu版本下]]></description><category></category></item><item><title><![CDATA[N3:(gensim库详解)直接用gensim 库训练word2vec模型]]></title><link>https://blog.youkuaiyun.com/weixin_57028107/article/details/130566790</link><guid>https://blog.youkuaiyun.com/weixin_57028107/article/details/130566790</guid><author>weixin_57028107</author><pubDate>Fri, 12 May 2023 10:50:27 +0800</pubDate><description><![CDATA[(gensim库详解)直接用gensim 库训练word2vec模型]]></description><category></category></item><item><title><![CDATA[N2中文文本分类项目：用transformer库中的bert实现文本分类]]></title><link>https://blog.youkuaiyun.com/weixin_57028107/article/details/130501887</link><guid>https://blog.youkuaiyun.com/weixin_57028107/article/details/130501887</guid><author>weixin_57028107</author><pubDate>Fri, 05 May 2023 16:24:13 +0800</pubDate><description><![CDATA[用bert实现多分类。注意点标签投影的设计。dataset的设计等]]></description><category></category></item><item><title><![CDATA[tokenizer,tokenizer.tokenize,tokenizer.encode,tokenizer.encode_plus它们到底有什么不一样？]]></title><link>https://blog.youkuaiyun.com/weixin_57028107/article/details/130506947</link><guid>https://blog.youkuaiyun.com/weixin_57028107/article/details/130506947</guid><author>weixin_57028107</author><pubDate>Fri, 05 May 2023 15:55:01 +0800</pubDate><description><![CDATA[第二种方式tokenizer.encode_plus(text)在分词后，还通过字典进行了词到id的映射 [101, 1045, 2066, 6207, 102]第一种方式tokenizer.encode(text)仅仅进行分词，把句子变为列表['i', 'like', 'apple']第三种方式不仅返回id列表，还以一个字典形式返回句子类型，注意力类型。最后推荐一直用encode_plus。第四种方式与上一个返回结果一样。]]></description><category></category></item><item><title><![CDATA[365天每周深度学习(R1)tensorflow实习RNN网络心脏病预测]]></title><link>https://blog.youkuaiyun.com/weixin_57028107/article/details/130393423</link><guid>https://blog.youkuaiyun.com/weixin_57028107/article/details/130393423</guid><author>weixin_57028107</author><pubDate>Wed, 26 Apr 2023 20:14:30 +0800</pubDate><description><![CDATA[tensorflow 写简单RNN模型]]></description><category></category></item><item><title><![CDATA[365天深度学习-自然语言处理-词嵌入加RNN学习（N1）]]></title><link>https://blog.youkuaiyun.com/weixin_57028107/article/details/130235192</link><guid>https://blog.youkuaiyun.com/weixin_57028107/article/details/130235192</guid><author>weixin_57028107</author><pubDate>Wed, 19 Apr 2023 16:47:36 +0800</pubDate><description><![CDATA[词嵌入技术+LSTM处理不定长序列]]></description><category></category></item><item><title><![CDATA[365 every week deep learning(t1)mnist num recognication based on tensorflow]]></title><link>https://blog.youkuaiyun.com/weixin_57028107/article/details/130019091</link><guid>https://blog.youkuaiyun.com/weixin_57028107/article/details/130019091</guid><author>weixin_57028107</author><pubDate>Tue, 11 Apr 2023 22:00:00 +0800</pubDate><description><![CDATA[第一周用tensorflow 实现mnist num recognication

主要学习：环境搭建、简单cnn网络搭建]]></description><category></category></item></channel></rss>