Text Clustering Algorithms

1. Hierarchical Agglomerative Clustering

in the very beginning, treat each input as a cluster, in each step, merge two closest clusters

2. K-means

build clusters around k centroids iteratively

define a good similarity measure (such as cosine measure based on the vector space model) is the key of these two algorithms

3. Suffix Tree Clustering

group inputs based on the identical phrases they share, two main phrases: discover base clusters and merge base clusters. the approach is difficult to tune and relevant inputs without common phrases will not be included in that cluster.

4. Semantic Online Hierarchical Clustering

similiar to 3, but, instead of using the whole sentence, a concept named "complete phrase" is defined, three main phrases: discover complete phrases and their frequencies in inputs, discover orthogonal base clusters using "single value decomposition", and merge base clusters into hierarchical structure.

LINGO is a variation of this approach. Most algorithms form clusters first and then figure out what labels to use. However, both LINGO (Carrot2) and Vivisimo try to find meaningful descriptions of clusters first, and determine their contents based on the descriptions.

Statistical Aspects of Data Mining

Lecture Videos

Other Interesting Videos:

Applied Machine Learning 2-Day Course
KDD 2007 Videos

Data Structures:

Dynamic Programming
What in Wikipedia?
Some Class Slides
Web Data Structures

A* Pathfinding for Beginners
How to Write a Spelling Corrector

(ZT) 隐马尔科夫模型HMM自学

(1) 介绍
(2) 马尔科夫模型也需要改进! (3) HMM 定义
(4-1) Forward Algorithm
(4-2) Forward Algorithm
(5-1) Viterbi Algorithm
(5-2) Viterbi Algorithm
(6) 尾声

如何防止HTML的表格因为内容太长而变得很宽？

用以下的CSS语句即可：

<style type="text/css">
    table
    {
        table-layout: fixed;
        word-wrap: break-word;
        overflow: hidden;
    }
</style>

统计学中的 t-test 浅述

首先，题外话说一个最简单的减肥口诀：早餐吃得像公主，晚餐吃得像乞丐。可惜我的意志力极端薄弱，每次都忍不住要吃。为了把自己推销出去，以及让自己过几年再沦为大妈，现在是时候严肃的对待体重问题乐。恐龙与美女的差距，就在于那10几磅肚腩肉，让我也跟着这个浅薄的世道一起浅薄吧。

突然懒得写了，大家看下面的自己揣摩吧：

(1) Student's t-test and Student's t-distribution
(2) Welch's t-test