Naive Bayesian Classification. Abel Sanchez, John R Williams. Stunningly Simple. The mathematics of Bayes Theorem are stunningly simple. In its most basic ...

Alex2发布于2018/06/09 00:00

注脚

1.Naive Bayesian Classification Abel Sanchez, John R Williams

2.Stunningly Simple The mathematics of Bayes Theorem are stunningly simple. In its most basic form, it is just an equation with three known variables and one unknown one. This simple formula can lead to surprising predictive insights.

3.Bayes and Laplace The intimate connection between probability, prediction, and scientific progress was thus well understood by Bayes and Laplace in the eighteenth century— the period when human societies were beginning to take the explosion of information that had become available with the invention of the printing press several centuries earlier, and finally translate it into sustained scientific, technological, and economic progress.

4.Conditional Probability Bayes’s theorem is concerned with conditional probability. That is, it tells us the probability that a hypothesis is true if some event has happened.

5.Bayes Theorem P(A) and P(B) are the probabilities of A and B independent of each other P(A|B) a conditional probability, is the probability of A given that B is true P(B|A) is the probability of B given that A is true

6.Example

7.Probability that your partner is cheating on you, given an event Event: you come home from a business trip to discover a strange pair of underwear. Condition: you have found the underwear Hypothesis: probability that you are being cheated on

8.p( u/c ) - The probability of underwear u given cheating c Probability of underwear appearing, conditional on his cheating 50%

9.p(u) - The probability of the underwear u appearing if NO cheating Probability of the underwear’s appearing conditional on the hypothesis being false . 5%

10.p(u ) - The probability of cheating c What is the probability you would have assigned to him cheating on you before you found the underwear? 4 %

11.p(u ) - The probability of cheating c What is the probability you would have assigned to him cheating on you before you found the underwear? 4 %

12.Active Learning

13.Active Learning – Calculate Cheating Probability T he probability of cheating c given underwear u The probability of underwear u given cheating c 50 The probability of the cheating c 4 The probability of the underwear u appearing if NO cheating 5

14.Classification of Drew

15.Example: classification of Drew We have two classes: c 1 =male, and c 2 =female Classifying drew as male or female is equivalent to asking is it more probable that drew is male or female. probability of being called “drew” given that you are a male? Probability of being a male? probability of being named “drew”?

16.Using Data Name Gender Drew Male Claudia Female Drew Female Drew Female Alberto Male Karin Female Nina Female Sergio Male p(male|drew) = 1/3 x 3/8 = 0.125 3/8 3/8 p(female|drew) = 2/5 x 5/8 = 0.250 3/8 3/8 probability of being called “drew” given that you are a male? Probability of being a male? probability of being named “drew”?

17.Bayesian Approach Posterior p robability based on prior probability plus a new event

18.Classification of Documents

19.Questions We Can Answer Is this spam? Who wrote which Federalist papers ? Positive or negative movie review ? What is the subject of this article?

20.Text Classification Assigning subject categories, topics, or genres Authorship identification Age/gender identification Language Identification Sentiment analysis …

21.Bayes Theorem a conditional probability, is the probability of class c given document d is the probability of document d given class c the probability of the class c the probability of the document d For a document d and a class c

22.The probability of a word given a class Count of the word occurring in that class Count of all words in that class Vocabulary – unique instances of words The probability of the class Number of documents with that class Total number of documents

23.Data Doc Words Class Training 1 chinese beijing chinese c Training 2 chinese chinese shanghai c Training 3 chinese macao c Training 4 tokyo japan chinese j Test 5 chinese chinese chinese tokyo japan ? Conditional Probabilities p(chinese| c ) = (5+1)/(8+6) = 6/14 p(tokyo| c ) = (0+1)/(8+6) = 1/14 p(japan| c ) = (0+1)/(8+6) = 1/14 p(chinese| j ) = (1+1)/(3+6) = 2/9 p(tokyo| j ) = (1+1)/(3+6) = 2/9 p(japan| j ) = (1+1)/(3+6) = 2/9 Priors p( c ) = 3/4 p( j ) = 1/4 Choosing a class (category) p(c| d5 ) = (3/4)*(3/7)*(3/7)*(3/7)*(1/14)*(1/14) ≈ 0.0003 p(j| d5 ) = (1/4)*(2/9)*(2/9)*(2/9)*(2/9)*(2/9) ≈ 0.0001

24.For homework we will use*: probability of language given word Probability that word is in language Probability that word is not in language * http://en.wikipedia.org/wiki/Naive_Bayes_spam_filtering

25.Calculating Probabilities // probability that word shows up in a language // probability that word is not in language

26.Underflow Prevention Multiplying lots of probabilities can result in floating-point underflow. Since log( xy ) = log(x) + log(y ); better to sum logs of probabilities instead of multiplying probabilities . Add probability of words (per language) using: In JavaScript ln is Math.log , and e is Math.exp At completion of each language:

user picture
  • Alex2
  • Apparently, this user prefers to keep an air of mystery about them.

相关Slides

  • 视觉任务之间是否有关系,或者它们是否无关?例如,表面法线可以简化估算图像的深度吗?直觉回答了这些问题,暗示了视觉任务中存在结构。了解这种结构具有显著的价值;它是传递学习的基本概念,并提供了一种原则性的方法来识别任务之间的冗余,例如,无缝地重用相关任务之间的监督或在一个系统中解决许多任务而不会增加复杂性。 我们提出了一种完全计算的方法来建模视觉任务的空间结构。这是通过在隐空间中的二十六个2D,2.5D,3D和语义任务的字典中查找(一阶和更高阶)传递学习依赖性来完成的。该产品是用于任务迁移学习的计算分类地图。我们研究了这种结构的后果,例如:非平凡的关系,并利用它们来减少对标签数据的需求。例如,我们表明,解决一组10个任务所需的标记数据点总数可以减少大约2/3(与独立训练相比),同时保持性能几乎相同。我们提供了一套用于计算和探测这种分类结构的工具,包括用户可以用来为其用例设计有效监督策略。

  • 尽管最近在生成图像建模方面取得了进展,但是从像ImageNet这样的复杂数据集中成功生成高分辨率,多样化的样本仍然是一个难以实现的目标。为此,我们以最大规模训练了生成性对抗网络,并研究了这种规模所特有的不稳定性。我们发现将正交正则化应用于生成器使得它适合于简单的“截断技巧”,允许通过截断潜在空间来精确控制样本保真度和多样性之间的权衡。我们的修改导致模型在类条件图像合成中达到了新的技术水平。当我们在ImageNet上以128×128分辨率进行训练时,我们的模型(BigGAN)的初始得分(IS)为166.3,Frechet初始距离(FID)为9.6,比之前的最优IS为52.52,FID为18.65有了显著的提升。

  • 2017年,以斯坦福大学为首、包括吴恩达、李开复等一众大咖专家团队齐力打造的人工智能指数(AI Index)重磅年度报告首次发布。从学术、业界发展、政府策略等方面对全年的人工智能全球发展进行了回顾,堪称全年人工智能最强报告。 该重点介绍了人工智能领域的投资和工作岗位前所未有的增长速度,尤其是在游戏和计算机视觉领域进展飞速。

  • 18年12月12日,哈佛大学,麻省理工学院,斯坦福大学以及OpenAI等联合发布了第二届人工智能指数(AI Index)年度报告。 人工智能领域这一行业的发展速度,不仅仅是通过实际产品的产生以及研究成果来衡量,还要考虑经济学家和政策制定者的预测和担忧。这个报告的目标是使用硬数据衡量人工智能领域的发展。 报告中多次提及了中国人工智能的发展以及清华大学: 美国仅占到全球论文发布内容的17%,欧洲是论文最高产的国家,18年发表的论文在全球范围内占比28%,中国紧随其后,占比25%。; 大学人工智能和机器学习相关课程注册率在全球范围都有大幅提升,其中最瞩目的是清华大学,相关课程2017年的注册率比2010年高出16倍,比2016年高出了将近3倍; 各国对人工智能应用方向重视不同。中国非常重视农业科学,工程和技术方面的应用,相比于2000年,2017年,中国加大了对农业方面的重视。 吴恩达也在今天的推特中重磅推荐了这份报告,称“数据太多了”,并划重点了两个报告亮点:人工智能在业界和学界都发展迅速;人工智能的发展仍需要更加多样包容。