java笔试题算法-hypergradient-descent:超梯度下降

上传者: 38623000 | 上传时间: 2023-01-23 16:24:37 | 文件大小: 17.83MB | 文件类型: ZIP
java笔试题算法超梯度下降 这是 ICLR 2018 论文的代码。 一个版本也在计划中,稍后会出现在这个 repo 中。 什么是“超梯度”? 在基于梯度的优化中,通过使用其关于模型参数的导数(梯度)来优化目标函数。 除了这个基本梯度之外,超梯度是相同目标函数相对于优化过程的超参数(例如学习率、动量或正则化参数)的导数。 可以有多种类型的超梯度,在这项工作中,我们对与标量学习率相关的超梯度感兴趣。 安装 pip install git+https://github.com/gbaydin/hypergradient-descent.git 我如何将它用于我的工作? 我们正在为 PyTorch 提供超梯度版本的 SGD(有或没有动量)和 Adam 优化器的现成实现。 这些符合torch.optim API,可用作代码中的直接替代品。 只需从这个 repo 中获取sgd_hd.py和adam_hd.py文件并像这样导入它们 from hypergrad import SGDHD , AdamHD ... optimizer = optim . AdamHD ( model . parame

文件下载

资源详情

[{"title":"( 37 个子文件 17.83MB ) java笔试题算法-hypergradient-descent:超梯度下降","children":[{"title":"hypergradient-descent-master","children":[{"title":"train.py <span style='color:#111;'> 11.50KB </span>","children":null,"spread":false},{"title":"results","children":[{"title":"vgg","children":[{"title":"+1e-03_+1e-03","children":[{"title":"sgdn.csv <span style='color:#111;'> 2.67MB </span>","children":null,"spread":false},{"title":"sgd.csv <span style='color:#111;'> 2.65MB </span>","children":null,"spread":false},{"title":"sgdn_hd.csv <span style='color:#111;'> 3.26MB </span>","children":null,"spread":false},{"title":"sgd_hd.csv <span style='color:#111;'> 3.24MB </span>","children":null,"spread":false}],"spread":true},{"title":"+1e-03_+1e-08","children":[{"title":"adam.csv <span style='color:#111;'> 2.68MB </span>","children":null,"spread":false},{"title":"adam_hd.csv <span style='color:#111;'> 3.30MB </span>","children":null,"spread":false}],"spread":true}],"spread":true},{"title":"logreg","children":[{"title":"+1e-03_+1e-07","children":[{"title":"adam.csv <span style='color:#111;'> 320.44KB </span>","children":null,"spread":false},{"title":"adam_hd.csv <span style='color:#111;'> 392.60KB </span>","children":null,"spread":false}],"spread":true},{"title":"+1e-03_+1e-03","children":[{"title":"sgdn.csv <span style='color:#111;'> 320.08KB </span>","children":null,"spread":false},{"title":"sgd.csv <span style='color:#111;'> 319.03KB </span>","children":null,"spread":false},{"title":"sgdn_hd.csv <span style='color:#111;'> 392.39KB </span>","children":null,"spread":false},{"title":"sgd_hd.csv <span style='color:#111;'> 389.99KB </span>","children":null,"spread":false}],"spread":true}],"spread":true},{"title":"mlp","children":[{"title":"+1e-03_+1e-07","children":[{"title":"adam.csv <span style='color:#111;'> 3.28MB </span>","children":null,"spread":false},{"title":"adam_hd.csv <span style='color:#111;'> 4.04MB </span>","children":null,"spread":false}],"spread":true},{"title":"+1e-03_+1e-03","children":[{"title":"sgdn.csv <span style='color:#111;'> 3.25MB </span>","children":null,"spread":false},{"title":"sgd.csv <span style='color:#111;'> 3.21MB </span>","children":null,"spread":false},{"title":"sgdn_hd.csv <span style='color:#111;'> 3.99MB </span>","children":null,"spread":false},{"title":"sgd_hd.csv <span style='color:#111;'> 3.93MB </span>","children":null,"spread":false}],"spread":true}],"spread":true}],"spread":true},{"title":"vgg.py <span style='color:#111;'> 2.89KB </span>","children":null,"spread":false},{"title":"run.sh <span style='color:#111;'> 1.80KB </span>","children":null,"spread":false},{"title":"poster","children":[{"title":"iclr_2018_poster.png <span style='color:#111;'> 1.10MB </span>","children":null,"spread":false},{"title":"iclr_2018_poster.pdf <span style='color:#111;'> 2.74MB </span>","children":null,"spread":false}],"spread":true},{"title":"hypergrad","children":[{"title":"adam_hd.py <span style='color:#111;'> 3.85KB </span>","children":null,"spread":false},{"title":"sgd_hd.py <span style='color:#111;'> 5.29KB </span>","children":null,"spread":false},{"title":"__init__.py <span style='color:#111;'> 54B </span>","children":null,"spread":false}],"spread":true},{"title":"plots","children":[{"title":"logreg.pdf <span style='color:#111;'> 104.29KB </span>","children":null,"spread":false},{"title":"vgg.png <span style='color:#111;'> 86.05KB </span>","children":null,"spread":false},{"title":"vgg.pdf <span style='color:#111;'> 146.01KB </span>","children":null,"spread":false},{"title":"mlp.pdf <span style='color:#111;'> 147.58KB </span>","children":null,"spread":false},{"title":"logreg.png <span style='color:#111;'> 63.85KB </span>","children":null,"spread":false},{"title":"mlp.png <span style='color:#111;'> 85.01KB </span>","children":null,"spread":false}],"spread":true},{"title":"LICENSE <span style='color:#111;'> 1.05KB </span>","children":null,"spread":false},{"title":"setup.py <span style='color:#111;'> 606B </span>","children":null,"spread":false},{"title":"README.md <span style='color:#111;'> 5.70KB </span>","children":null,"spread":false},{"title":".gitignore <span style='color:#111;'> 1.16KB </span>","children":null,"spread":false},{"title":"plot.py <span style='color:#111;'> 4.48KB </span>","children":null,"spread":false}],"spread":false}],"spread":true}]

评论信息

免责申明

【只为小站】的资源来自网友分享,仅供学习研究,请务必在下载后24小时内给予删除,不得用于其他任何用途,否则后果自负。基于互联网的特殊性,【只为小站】 无法对用户传输的作品、信息、内容的权属或合法性、合规性、真实性、科学性、完整权、有效性等进行实质审查;无论 【只为小站】 经营者是否已进行审查,用户均应自行承担因其传输的作品、信息、内容而可能或已经产生的侵权或权属纠纷等法律责任。
本站所有资源不代表本站的观点或立场,基于网友分享,根据中国法律《信息网络传播权保护条例》第二十二条之规定,若资源存在侵权或相关问题请联系本站客服人员,zhiweidada#qq.com,请把#换成@,本站将给予最大的支持与配合,做到及时反馈和处理。关于更多版权及免责申明参见 版权及免责申明