Yifan's Manuscript

Weakness and ignorance are not barriers to survival, but arrogance is.

笔记 | 半监督学习

周志华老师的《机器学习》一书第十三章读书笔记

本章主要关注数据中一部分已标注，一部分未标注的情形，一般记为，\ D_l = {(x_1, y_1), …, (x_l, y_l)}, D_u = {x_{l+1}, …, x_{l+u}}\ \( y_i \in \lbrace-1, +1\rbrace \)。 13.1 未标记样本一个简单的做法是主动学习：用 $D_l$ 先训练一个模型，拿这个模型去地里挑一个瓜，询问瓜农好...

Posted by Yifan on January 6, 2020

Review | Random Features for Large-Scale Kernel Machines

Goal Accelerate the computation of kernel matrix by directly approximating the kernel function. Map x into a low-dimension random feature space, $z: \mathcal { R } ^ { d } \rightarrow \mathcal { ...

Posted by Yifan on February 8, 2019

A Trip in Pennsylvania

Pittsburgh -> State College -> Philadelphia -> Washington

Day 2 @ CMU The signature sculpture, ‘Walking to the Sky’ Painted Fence Hunt Library The sculpture of Dr. Yisheng Mao Gates Center for Computer Science ...

Posted by Yifan on December 30, 2018

笔记 | 计算学习理论

周志华老师的《机器学习》一书第十二章读书笔记

本章主要讨论二分类问题，若无特别说明，\( y_i \in \lbrace-1, +1\rbrace \)。 12.1 基础知识泛化误差：E(h), 预测不符的概率经验误差：\( \hat{E}(h) \), 数据集上预测不符的频率几个常用不等式： Jensen 不等式：期望的凸函数 f(E) <= 凸函数的期...

Posted by Yifan on August 10, 2018

A Convenient Image Compression Tool -- Luban-Py

Very Close to the Efficacy of WeChat

Need for Convenient Compression Recently I began to operate my own blog, and I found a high-quality image in my blog is attractive, however it will take users much time to load it. In that case I ...

Posted by Yifan on August 9, 2018

Github Pages 个人主页的配置

关于本地调试与数学公式支持

感谢Huxpro提供的博客模板感谢BY的详细指导本文的环境为 Windows 10 前置步骤本文搭建的博客主要利用了 Github 提供的 Github Pages 服务，因此事先需要注册一个 Github 账号，并完成一系列配置。这一部分 BY 前辈写的比较详细了，可以直接参考他的博客中“快速开始”部分。本地调试本文主要介绍下在当前时间节点下（2018-07...

Posted by Yifan on August 1, 2018