Abstract:

Conjugate gradient method is one of the improved gradient algorithms for optimization.
It’s used to resolve linear system, which need to meet some strict requirements.
Though it’s not a general solution for optimization problem, but it’s quite efficient and memory saving for problems it fit.

Abstract:

Main page for recording learning pandas based on book “Python for Data Analysis” by Wes McKinney and his example codes from repos.
This book introduced basic ipython tool, bases numpy and then pandas. It shows a easy way to process 2-ndarray as a df. The example notebook include my own notes when learning or test. Enjoy it!

Abstract:

Main page for recording learning sklearn from sklearn web.
It include many machine learning algorithm/application, both supervised or unsupervised learning, such as classification, regression, clustering, etc. Enjoy it!

Abstract:

I demostrate how I highlight the code in my jekyll v3.* post by rouge.

Abstract:

To copy a file or dir is useful, python provide it by shutil, there are functions copyfile, copy, copymode, copy2, copytree.
This pages note the differences between them.

Abstract:

This program is to test python process pool.

Abstract:

之前介绍了用PC机借助虚拟机搭建分布式hadoop，这篇介绍用它体验hadoop的wordcount操作。

hadoop上最经典的入门案例就是wordcount了，经典到在hadoop2.5.2的安装包里默认装了wordcount的jar包。所以我们免除了写java文件的工作(有兴趣的话可以研究它的源码，从而模仿改写)，直接调用体验。 不过由于是在单机上虚拟出来的分布式环境，其实硬件资源跑三个虚拟机已经够呛了，所以hadoop的威力体现并不明显，wordcount跑了不少时间。但至少成功体验了跑一遍job，对hadoop感受更直接了。

Abstract:

numpy和scipy是python的第三方模块，但是数学计算(ML)中很有用。由于不是内置的，使用前需要先安装，在16.04下装这些python模块简直不要太简单。

之前一直坚持使用的ubuntu10.04，由于更新库确实不能再用了，一咬牙直接更新到了16.04(目前最新版)。使用新版的好处就是源多，而且都是基本最新的，在16.04下装这些python模块简直不要太简单。

Abstract:

用VirtualBox装了３个ubuntu10.04的32位系统，并在该环境下装了hadoop-2.5.2。

Abstract:

Singleton是创建模式中很典型的一类，在Java中是通过将构造函数private来实现的。在python中，没有private，我们通过share一个类或子类的所有成员变量来实现单例。注意，虽然所有对象都共享相同的成员，但和Java中的单例还是有差别。java中是真的只建了一个对象，但python中实际有多个对象(id不同), 但成员相同。