SQL Summary

Creating Tables

1
2
3
4
5
6
7
8

create table unit
(
    unit_code  char(7)  not null,
    unit_name  varchar2(50) constraint uq_unit_name unique not null,
    student_id number(7, 0)  not null,
    constraint pk_unit primary key (unit_code)
    constraint fk_student foreign key (student_id) references student(student_id)
);

梯度下降

梯度下降是机器学习优化中常用的算法，因为凸函数 (convex function) 沿着负梯度方向移动一定能够到达全局最小值 (global minimum)，在实际应用中由于损失函数大多数情况下不能保证为凸函数，因此梯度下降法更可能在局部 (局部凸函数) 找到局部极小值 (local minimum).

MAT9004复习总结

快到期末了......最近在复习MAT9004这门课，这门课主要讲的是高数、线性代数和概率论的一些基础知识，lecture上一般讲的比较简单，但每个知识点的切入点都很不错，在复习过程中有一些困惑，结合一些资料以及之前的知识储备浅略的探索了一下，总结的可能比较零碎，如果篇幅较长的话后期再做分类整理.

Shell Summary ## Test

Auto Encoder

Auto encoder is a kind of method to reduce the dimension of input data, it's a shallow neural network which learning the weights by back propagation.

隐语义模型

关于隐语义模型(latent factor model)的介绍，项亮博士在《推荐系统实践》中有过介绍，这本书的确是推荐系统方向非常非常棒的书籍. 我在用python实现书中隐语义模型时遇到了一些问题和坑，因此用这篇博客进行记录并总结，之后有新的或更深的理解会及时更新本文.

2019.5.2 近期问题总结

pandas 过滤数据后比较

pandas 使用 loc()方法过滤(获取)到的数据不能够直接在if, while, for等语句中和其它常用类型直接比较.

Regularization in Machine Learning

About Regularization

The model will become overfitting when it is trying too hard to capture the outliers in the training dataset and the accuracy will have a low accuarcy in the testing set.

ROC & AUC

Confusion Matrix

I am going to introduce the concept of Confusion Matrix before talking about the ROC as well as AUC since I think this will help us to understand them better. Confusion Matrix is a performance measurement for machine learning classification, which is a specific table layout. Let us undetstand TP, FP, FN and TN first.