Coursera Game theory experience

I had no past foundation in game theory before taking this course. I am an undergrad material science major with the relating numerical skill (up to multivariable analytics, differential conditions, and so forth.), and a starting course in Discrete Math. So I had seen the set theory and summation documentation previously, however, had no experience applying them to game theory. 

The educators were incredible and unmistakably have a profound scholastic foundation in the subject, just as a noteworthy individual intrigue. This is significant in light of the fact that they didn't appear to be on the off chance that they were simply handing-off data; they gave the impression of really getting a charge out of the material and in any event, going into brief intervals about the inspiration driving a portion of the definitions and models from their own points of view, notwithstanding the more "standard" address material. 

The course is unquestionably a presentation, and doesn't go into the greater part of the more proper evidence of the hypotheses utilized, nor does it use models or schoolwork issues that require a critical foundation in the subject. I discovered I had the option to finish the greater part of the issue sets inside 4-5 hours, with a normal of around 3 hours. Be that as it may, your mileage on this may change, particularly in case you're less acquainted with the numerical documentation and the style of critical thinking utilized. This isn't to state the issues are not troublesome, on the grounds that they are! A large number of them are non-instinctive and expect you to think around corners, or think about strategies for speculation and critical thinking that are not ordinarily utilized in math and the common sciences. This, I believe, is a result of game theory's prevalent figure - John Nash's - exceptional way to deal with arithmetic. He was celebrated for utilizing exceptionally natural, non-thorough methods for reaching resolutions before demonstrating them all the more officially. In this way, the field is somewhat unpredictable in its techniques and approach, and won't be consequently open since somebody has just had an introduction to cutting edge math or science. It requires time and cautious idea to build up a profound comprehension of. In any case, I found that this exertion was very much compensated before the finish of the course where I began to see all the prior hypotheses and procedures meeting up to shape a bound together arrangement of critical thinking capacities. 

I emphatically expect to take the propelled rendition of this course is likewise offered on Coursera. Concerning this one, I prescribe it to anybody keen on understanding the science of games, serious situations, and complex frameworks by and large! Done by the teachers.

Machine Learning course experience on Coursera

Of late I finished my Machine Learning course by Stanford University on Coursera. I figured it may be useful to share my experience quickly. I picked Machine Learning since I took a few courses around this point in college years back. Consequently, I previously had a few (nearly overlooked) information and needed to utilize my specialized foundation more once more.

Contents

The course itself is organized quite well. The whole substance is separated into eleven weeks, each with up to four sub-units. Among others, the course covers the accompanying subjects: 

upervised learning (e.g. linear regression, logistic regression, neural networks, SVM)
Unsupervised learning (e.g. K-means, PCA, anomaly detection)
Fancy topics (e.g. recommender systems, OCR, MapReduce)
Analysis & Evaluation (e.g. learning curves, error analysis, ceiling analysis)

The instructor Andrew NG presents the material in video instructional exercises. Exactly he clarifies the substance utilizing PowerPoint slides and imprint ups to compose a lot of extra data in the slides. There is no compelling reason to record anything, since all slides including the imprint ups are accessible for download.

Assessment

Toward the finish of most sub-units, there are short tests with five inquiries each. The inquiries extend from different decisions over single-decision to correct outcome computation. I found these tests the hardest piece of the course. Toward the finish of most weeks, there is a mind-boggling programming task, which must be executed in either MATLAB or Octave (open source). I didn't think the programming practices are convoluted. The test was to peruse the directions cautiously and approach just slowly and carefully.

Conclusion:

The course is very tedious. Completing it inside eleven weeks, as suggested by Coursera, requires devoting around balanced and a half-day to the course every week. Towards the end, the exertion turns out to be less. Except if you need to get the last declaration, the course is free. I can energetically prescribe Coursera as a learning stage and the course Machine Learning by Andrew NG specifically.

武漢肺炎終極地圖

武漢肺炎終極地圖

http://virus-trend.com/

介紹兩個 unsupervised learning 演算法 PCA 跟 diffusion map

介紹兩個 unsupervised learning 演算法 PCA 跟 diffusion map

兩者非常像
基本上就是原資料生出某矩陣
計算eigenvalue 跟 eigenvector即可

因為一開始的矩陣決定了你的特徵向量代表的物理意義


只找前幾個的原因是這些算法都在尋找最佳表示法


而線性代數中對角化就是在找對佳表示法


我講最大的區別在哪裡




最大的區別就是PCA在找的最佳表示法是線性的

擴散圖是非線性的


PCA利用的是一般的平移旋轉


diffusion map用的是熱統計的擴散方程的解


它假設一個數據點到另一個數據點用擴散的連過去