讀機器學習方向的博士哪個學校最好？斯坦福，麻省理工，伯克利還是卡內基梅隆？

01-11

這個問題是很大，但是還是很有意思的。
可以從各個學校的professor或者主要研究的領域漫談吧

根據我的了解，簡單介紹一下四校（按校名排序）機器學習的情況（可能會有很多不準確的地方）。

Berkeley：小而精。從事機器學習研究的教授不多，但是有大牛Michael Jordan和Martin Wainwright 坐鎮，整體實力非常強。最近新去了年輕的明星教授Ben Recht（之前在威斯康星）。

CMU：規模大，整體實力非常強。專門設置了一個Machine Learning Department，彙集了十幾位機器學習教授。偏理論的有（按姓排序）Nina Balcan，Avrim Blum, Barnabás Póczos，Aarti Singh，Larry Wasserman, 偏模型和演算法的有Geoffrey Gordon，Jeff Schneider，Alex Smola，Ryan Tibshirani， Eric Xing，Yiming Yang，偏應用的有Ziv Bar-Joseph，William Cohen, Christos Faloutsos, Tom Mitchell。（註：上述劃分並非絕對，很多教授從理論，到模型，到應用均有涉及）。

MIT：長期以來，MIT不重視機器學習，基本上是靠Tommi Jaakkola一人支持。2014年突然招了Michael Jordan的兩個學生，Tamara Broderick和Stefanie Jegelka，開始大力發展ML。另外，認知科學系的Josh Tenenbaum教授也做一些機器學習，做的非常好。

Stanford：曾經實力非常強，有大牛Andrew Ng和Daphne Kollar。後來隨著兩位教授興趣和職業方向的轉變，Stanford ML下滑。最近兩年新招了兩位新銳：Percy Liang和John Duchi (均為Michael Jordan的學生），又開始迅速上升。

給橫掃四個學校的大神跪了

機器學習方向都很厲害，沒有什麼第一或第二。如果你想真要排名，你先要弄清楚具體哪個方向，機器學習領域很大，包含很多子領域。

有這個實力的人不至於不會自己找資料，而且也不會用這種老子隨便挑的口氣問。你好好努力，錄啥去啥吧。

反對一下「錄你的最好」這個答案。

退一步講我們考不上清華北大就不能浪費我們「自己」的時間去討論一下清華和北大的哪個師資好了么？

要是有個人問一個問題：林志玲、林心如、林依晨哪個更漂亮？是不是改回答，願意嫁給你的最好。不嫁給我我就不准許欣賞了么？欣賞完我還不能默默打個分么？

橫掃大神Orz

有種從小糾結上清華還是北大的既視感

也許你需要這個幫助決定

結論：都是喬丹一夥兒的。。。

弱弱問一句，你是可以隨便挑了嗎？

如果是，這選擇還真是頭疼嚯。

machine learning來說。Berkeley和cmu應該是首選吧。Berkeley的stat應該是全北美最強的了。

另外可以考慮下歐洲幾所學校，Cambridge的zoubin組以及Amsterdam的max welling組，也是強到爆炸。

真心最喜歡喜歡Cambridge的ml組，可以看看各個組phd的thesis。

同上，機器學習方面的都很厲害

哪怕在我們學校教machine learning的，也都是美帝一流名校出身

關鍵在於自己

建議斯坦福或者CMU

CMU似乎不是特別卡GPA

Carnegie Mellon managed to beat Google by just 1 paper. Microsoft and Stanford also managed to publish more than 80 papers in 2016. IBM, Cambridge, Washington and MIT all reached the 50 publication barrier. Google, Stanford, MIT and Princeton are distinctly focused on the ML aspect, publishing mostly in NIPS and ICML. In fact, Google papers counted for nearly 10% of all NIPS papers. IBM, Peking, Edinburgh and Darmstadt however are distinctly focused on the NLP applications.

數據來源：

人工智慧的頂級會議和期刊

includes 11 different conferences and journals: ACL, EACL, NAACL, EMNLP, COLING, CL, TACL, CoNLL, *Sem/StarSem, NIPS, and ICML.

其實類似的疑惑我也有很多

（我是上清華好還是上交好

當教授好還是當CEO好

去美國的四大好還是英國的牛津劍橋好

娶官二代好還是富二代好

留美國好還是留日本好

去華爾街好還是矽谷好

做金融好還是做互聯網好）

@vczh

What are the best graduate schools for studying machine learning?

With the possible exception of CMU (which has a machine learning department), the answer really depends on which professors at each school are currently research active and open to taking on new students. Most schools only have a handful of professors that work in ML-related areas, so prior reputation of a school is not 100% indicative that you"ll find opportunities there. For example, a school like Stanford is overall extremely strong, but many of its core ML professors are currently very busy with startups (e.g., Coursera). In my opinion, the most important thing in doing a PhD is to have a good fit with your advisor.
Roughly speaking, I would say that most of the schools in the top 35 (as ranked by US News) have 8-12 professors actively working in ML-related areas, and 1-4 in core machine learning (i.e., regularly publishing in ICML, NIPS, KDD, etc). I also wouldn"t blindly trust the faculty listings on school websites -- those are usually inflated.
CMU is about an order of magnitude bigger with about 8-12 faculty in core ML (the machine learning department at CMU is actually pretty small, so the faculty count is not as ginormous as some might think). So CMU is more robust to individual faculty leaving or not being research active due to a startup. But there are also more students there interested in ML so that leads to more competition for faculty advising.
Currently (ca 2014), some schools at the high end of the 1-4 core ML faculty are Michigan, Berkeley, Stanford (mostly related to vision NLP), Columbia, UWashington, GA Tech, Toronto, and UCSD. I would also say that UCLA, UNC, and Rice currently have a clear dearth of ML professors (at least based on my observations at ICML, NIPS, KDD, etc).
Finally, keep in mind that different professors work on very different topics, even if it"s all under the umbrella of machine learning. Research universities try to hire professors that are doing both interesting and novel work, so each professor"s research profile is unique to some degree. I hear that Caltech has some interesting professors