blog.huajh7.com
Road2beauty : The road to a beautiful world.
http://www.blog.huajh7.com/road-2-beauty
Imagination is more important than knowledge. -Albert Einstein. The road to a beautiful world. 2) Mean field Method.
blog.huajh7.com
A Beautiful Mind Archives - Junhao Hua
http://www.blog.huajh7.com/category/beautifulmind
Imagination is more important than knowledge. -Albert Einstein. April 1, 2013. 我不清楚是否已经后悔说 更愿意直博 ，不知道当初怎么脑抽，觉得自己应当走 理论 研究这条路。 每个人心中多多少少都会有一个 Role model 我指的是Albert Einstein, Michelangelo ，他在某些方面给人以动力，目标，准则。 然而，充满 欲望 与 信念 的你，肯定不会停止脚步，肯定不会让自己大脑停留下空白的虚无之中。 你厌恶唯 paper论，讨厌为一点点所谓的 Paper创新 而东拼西凑的知识零件，这并不是对交叉学科的偏见，而是对 因为paper 所以paper 的憎恨。 你有让你的家人活的更有尊严，活得更有价值的义务 你有让你的家族变得更好的义务 你无法抗拒来自社会的偏见与傲慢 无论采取何种可被接受的方式，你都得征服它。 诚然, 美好 的含义是你所定义的，千万不能落入庸俗，被社会所谓的价值观所淹没。 July 19, 2011. 3) Mean field Method. 1) Mean field Method.
blog.huajh7.com
Machine Learning Archives - Junhao Hua
http://www.blog.huajh7.com/category/machinelearning
Imagination is more important than knowledge. -Albert Einstein. Coursera : PGMs & ML Lecture Notes. April 19, 2014. 1 Probabilistic Graphical Models (PGMs) - Daphne Koller From Apr 8th - Jun 24th, 2013 (11 weeks long) Workload: 15-20 hours/week 主讲人 koller是来自以色列的神童级人物，请自行wiki, 有些同学感觉koller的英文不好懂，说得比较晦涩。 作业 PGM的作业，特别是编程作业是比较难的，每周基本需要花费20小时左右的时间 不包括吃饭放松休息的时间 ，有几周，如果一个问题卡住了，一不小心花费30小时，也是很正常的。 PA Quiz Final : 0.98027 0.95299 0.672 总成绩(with distinction) 91.6% 笔记 课程笔记本来想写成blog的，一直没有时间写。 April 18, 2014. 这里先给几个较小...
blog.huajh7.com
lecture Archives - Junhao Hua
http://www.blog.huajh7.com/tag/lecture
Imagination is more important than knowledge. -Albert Einstein. Coursera : PGMs & ML Lecture Notes. April 19, 2014. 1 Probabilistic Graphical Models (PGMs) - Daphne Koller From Apr 8th - Jun 24th, 2013 (11 weeks long) Workload: 15-20 hours/week 主讲人 koller是来自以色列的神童级人物，请自行wiki, 有些同学感觉koller的英文不好懂，说得比较晦涩。 作业 PGM的作业，特别是编程作业是比较难的，每周基本需要花费20小时左右的时间 不包括吃饭放松休息的时间 ，有几周，如果一个问题卡住了，一不小心花费30小时，也是很正常的。 PA Quiz Final : 0.98027 0.95299 0.672 总成绩(with distinction) 91.6% 笔记 课程笔记本来想写成blog的，一直没有时间写。 2) Mean field Method.
blog.huajh7.com
coursera Archives - Junhao Hua
http://www.blog.huajh7.com/tag/coursera
Imagination is more important than knowledge. -Albert Einstein. Coursera : PGMs & ML Lecture Notes. April 19, 2014. 1 Probabilistic Graphical Models (PGMs) - Daphne Koller From Apr 8th - Jun 24th, 2013 (11 weeks long) Workload: 15-20 hours/week 主讲人 koller是来自以色列的神童级人物，请自行wiki, 有些同学感觉koller的英文不好懂，说得比较晦涩。 作业 PGM的作业，特别是编程作业是比较难的，每周基本需要花费20小时左右的时间 不包括吃饭放松休息的时间 ，有几周，如果一个问题卡住了，一不小心花费30小时，也是很正常的。 PA Quiz Final : 0.98027 0.95299 0.672 总成绩(with distinction) 91.6% 笔记 课程笔记本来想写成blog的，一直没有时间写。 1) Mean field Method.
blog.huajh7.com
Mathematics Archives - Junhao Hua
http://www.blog.huajh7.com/category/mathematics
Imagination is more important than knowledge. -Albert Einstein. A Cauchy matrix is infinitely divisible. November 10, 2014. This article gives a proof of Example 4.11 in the textbook: Matrix Analysis with some Application. It shows that a Cauchy matrix is infinitely divisible. Theorem 1. Take numbers (r, lambda 1, lambda 2, dots, lambda n 0 ). And set (A {ij} = frac{1}{( lambda i lambda j) r} (1 leq i,j leq n), ). The matrix (A ). Is positive semidefinite for all (r 0 ). Bayesian MSE vs. Classical MSE.
blog.huajh7.com
A Cauchy matrix is infinitely divisible - Junhao Hua
http://www.blog.huajh7.com/cauchy-matrix-infinitely-divisible
Imagination is more important than knowledge. -Albert Einstein. A Cauchy matrix is infinitely divisible. A Cauchy matrix is infinitely divisible. November 10, 2014. This article gives a proof of Example 4.11 in the textbook: Matrix. Analysis with some Application. It shows that a Cauchy matrix is. Take numbers (r, lambda 1, lambda 2, dots, lambda n 0 ). A {ij} = frac{1}{( lambda i lambda j) r} (1 leq i,j leq n), ). The matrix (A ). Is positive semidefinite for all (r 0 ). Article written by huajh7.
blog.huajh7.com
Variational Inference Archives - Junhao Hua
http://www.blog.huajh7.com/category/machinelearning/variational-inference
Imagination is more important than knowledge. -Albert Einstein. Talk at csmath]: A Tutorial on Variational Bayes. April 18, 2014. 虽然我花费了近一节课时间 本来力求控制在15-20分钟 ，但我相信90% 的同学听完后依然是一头雾水或者中途便放弃了。 课后与之聊了几句后，此人竟继续讲了PGM 当然，我相信是因为他之前正好也看过的那本书(Wainwright, Jordan，2008) ，甚为惊喜。 其实呢，从对听众负责任的角度讲，为了让其有所收获，或许不需要有多么厉害，多么酷的内容，而在于如何引着听众朋友让其感受到你讲的故事是多么美好，让其有所思，有所悟，并激发其潜在的渴望: 我回去一定要读一读，我现在就在ipad搜了一堆在读了，真有意思 SLIDES: csmath vb Talk.[PDF]. March 23, 2013. 11 理论基础 Bayesian networks 变分信息传播方法是建立在贝叶斯网络[3]上的，如图所示，对于一个节点 ({H j} ).
blog.huajh7.com
Coursera : PGMs & ML Lecture Notes - Junhao Hua
http://www.blog.huajh7.com/lecture_note_pgm_ml_2013
Imagination is more important than knowledge. -Albert Einstein. Coursera : PGMs & ML Lecture Notes. Coursera : PGMs & ML Lecture Notes. April 19, 2014. 1 Probabilistic Graphical Models (PGMs). From Apr 8th - Jun 24th, 2013 (11 weeks long). PA Quiz Final : 0.98027 0.95299 0.672 总成绩(with distinction) 91.6%. 笔记分享 https:/ www.evernote.com/pub/huajh/pgm-koller. From Apr 22nd - Jul 1st, 2013 (10 weeks long). 所以最终成绩是without distinction，得了 97.5%. Https:/ www.evernote.com/pub/huajh/machinelearing-andrewng.
blog.huajh7.com
Reading List - Junhao Hua
http://www.blog.huajh7.com/road-2-beauty/readinglist
Imagination is more important than knowledge. -Albert Einstein. Peter J. Schreier. Statistical signal processing of complex-valued data: the theory of improper and noncircular signals. CM Bishop, Pattern Recogntion and Machine Learning. Probabilistic graphical models: principles and techniques,. Roger A. Horn,. Kevin P. Murphy. Machine Learning - A Probabilistic Perspective. James V. Candy,. B ayesian Signal Processing: Classical, Modern and Particle Filtering Methods. Thomas M. Cover. David J C MacKay，.
