Machine Learning

[url]https://www.ibm.com/developerworks/java/library/j-mahout/index.html[/url][quote]Machine learning is a subfield of artificial intelligence concerned with techniques that allow computers to improve their outputs based on previous experiences. The field is closely related to data mining and often uses techniques from statistics, probability theory, pattern recognition, and a host of other areas.
Supervised learning is tasked with learning a function from labeled training data in order to predict the value of any valid input. Common examples of supervised learning include classifying e-mail messages as spam, labeling Web pages according to their genre, and recognizing handwriting. Many algorithms are used to create supervised learners, the most common being neural networks, Support Vector Machines (SVMs), and Naive Bayes classifiers.
Unsupervised learning, as you might guess, is tasked with making sense of data without any examples of what is correct or incorrect. It is most commonly used for clustering similar input into logical groups. It also can be used to reduce the number of dimensions in a data set in order to focus on only the most useful attributes, or to detect trends. Common approaches to unsupervised learning include k-Means, hierarchical clustering, and self-organizing maps.[/quote]

ppt:
[url]http://www.slideshare.net/Cataldo/tutoria-mahout-recommendation[/url]
[url]http://www.slideshare.net/JeanPierreKnig/what-are-product-recommendations-and-how-do-they-work[/url]
[url]http://www.slideshare.net/erikbern/collaborative-filtering-at-spotify-16182818[/url]
text:
[url]http://www.bidn.com/blogs/cprice1979/ssas/4388/-mahout-recommendation-engines-part-2-ride-the-elephant[/url]


[b]Algorithms：[/b]
[url]https://cwiki.apache.org/confluence/display/MAHOUT/Algorithms[/url]
[url]http://answers.google.com/answers/main?cmd=threadview&id=225316[/url]
LSH (Locality-sensitive hashing)
SimHash http://my.oschina.net/pathenon/blog/63747
MinHash http://my.oschina.net/pathenon/blog/65210
TF-IDF & VSM http://pyevolve.sourceforge.net/wordpress/?p=2497


Clustering - unsupervised learning - 不知道什么结果
Classification - supervised learning - 有已知的固定结果集


[b]Machine Learning Terms：[/b]
Association Rules - 关联规则，一个消费者购买了产品A，那么他有多大机会购买产品B?
latent semantic analysis - LSA
probabilistic latent semantic analysis - pLSA
Latent Dirichlet Allocation - LDA [url]http://blog.echen.me/2011/08/22/introduction-to-latent-dirichlet-allocation/[/url]


[b]Math Terms:[/b]
Derivative - 导数
covariance - 协方差 常缩写为 cov
standard deviation - 标准差 常缩写为 stddev
dot product - 数量积（也称为内积、标量积、点积、点乘）


[b]Math symbols：[/b]
[url]http://en.wikipedia.org/wiki/List_of_mathematical_symbols[/url]
[url]http://zh.wikipedia.org/wiki/%E7%94%A8%E6%96%BC%E6%95%B8%E5%AD%B8%E3%80%81%E7%A7%91%E5%AD%B8%E5%92%8C%E5%B7%A5%E7%A8%8B%E7%9A%84%E5%B8%8C%E8%87%98%E5%AD%97%E6%AF%8D[/url]


[b]Text processing:[/b]
Feature Selection，Feature Extraction, Terminology Extraction，Keyword Extraction, 有什么不同？
这里介绍了 Feature Extraction 和 Feature Selection 的区别：
[url]http://stackoverflow.com/questions/2163330/difference-between-feature-selection-feature-extraction-feature-weights[/url]
至于 Feature Extraction / Terminology Extraction / Keyword Extraction 这三个词在含义上有没有什么本质的区别，我tm也不知道，只知道 wikipedia 上FE和TE两个条目是独立的。
Collocation - 搭配词
n-gram - a contiguous sequence of n items from a given sequence of text.
Stemming - process for reducing inflected (or sometimes derived) words to their stem, base or root form—generally a written word form. 即 cats catty ... 都统一为 cat。


[b]Evaluation of recommendation:[/b]
有 preferences 存在的，可以使用 rms root mean squared 等来做；
对于 没有 preferences 的 boolean 推荐，怎么评估那？理论上可以用 classic Information Retrieval metrics：Precision & Recall 来做；但是要注意，基于 Precision & Recall 来做对推荐结果的评估并不是理想的方案。见 Sean：
[url]http://lucene.472066.n3.nabble.com/Evaluating-Boolean-Preferences-for-Item-Recommenders-td688560.html[/url][quote]I am not surprised by low precision. It doesn't necessarily mean the recommender is bad (though it could!). I think a precision-recall test is somewhat flawed for recommenders. It measures how well the recommender returns things the user has already seen, which are not necessarily the best recommendations. That is, a recommender gets penalized in this test if it recommends something the user *would* like, but hasn't rated. [/quote][url]http://stackoverflow.com/questions/7529333/similarity-function-for-mahout-boolean-user-based-recommender[/url]
关于 mahout 实现的 IR metrics 中的 Precision / Recall / Fall-out / nDCG 等：
[url]http://en.wikipedia.org/wiki/Information_retrieval[/url]
[url]http://stackoverflow.com/questions/16478192/how-to-interpret-irstatisticsimpl-data-in-mahout[/url]