Duke MIDS

[ds2207]

I am really confused about Duke's MIDS program. It is new but Duke is a big name.

Is it a good program and worth attending for masters in data science

Edited April 4, 2020 by ds2207

[bayessays]

I'm just a guy on the internet who has been reading the websites of some of these programs lately, so I don't have any first have knowledge of these programs.  But I have worked as a data scientist at top companies and have some knowledge of what would make a good program. Some data science programs seem like decent options that allow you to take a mix of stats/CS courses (like Penn's program) and some seem like impressive career-oriented preparation (Northwestern's analytics program).

Given the high cost of Duke's program and the syllabi of courses, to me it looks like a very expensive long bootcamp program (like Metis, Galvanize).  To me, the classes look very superficial and you would probably come out of the program with decent Python knowledge, a very basic high level understanding of some statistics/ML concepts, and a project. I personally don't think that is worth over $100,000.

[dontoverfit]

I would strongly recommend you do not attend MIDS. It can be hard figuring out how legitimate some of these Data Science programs are, and unfortunately MIDS is one of the ones that just is not worth the cost. Like bayessays said above, it is essentially a long bootcamp. Classes are extremely basic and are not taught by some of the top data science faculty at Duke. However, some of the students have gone on to get good internships/jobs (although it's so new that there isn't a long track record) partially because (1) Duke's name is good and (2) several of the students were quite knowledgeable already. If you're just looking for a school name but care little about the quality of the program itself, it's an okay but very pricey option. Let's put a few things into perspective....

(1) The courses are "special" data science classes. This is often concerning from some of the data science programs. Students don't take, as core classes, classes from statistics, mathematics, or computer science because they are not well enough prepared to handle the rigor of the other classes. I know of students who have taken some of those classes as electives who have had to drop them because they couldn't keep up. And the statistics department wants essentially nothing to do with this program...most of the core masters level statistics classes are closed off to MIDS students because it would slow down the class.

(2) There is so much turmoil in the leadership of the program. Last year, a professor quit early in the semester. The leadership has changed so much. Many of the courses are taught by professors who do not want to be very involved in the program. There has been so much upheaval, and while the program is young, that is not a great sign.

 

But has some much much stronger MS programs for data science. In particular, I would check out the statistical science degree, the economics/cs dual degree, the CS degree, or some of the ECE degrees if you would like to get a solid education. I genuinely feel bad for the students in the program because it is such a mess.

[MIDS_Student]

I wanted to reply to this post as what dontoverfit says in his comment is in many aspects false and a quite damaging. I am a current MIDS student so I am not partial about the program, but it does not seem dontoverfit is partial either.

 

The program did go through a lot of turmoil in the leadership team during the past year, with the director leaving. I don't know of any teachers that quit, but some courses changed for the next year, given reviews by students. I believe that was for the best and while it was a bit too much, it is still a new program. Since then the program has stabilized a lot, there haven't been any major changes for over a year and I don't expect the program to change much anymore.

 

The program has 'specialized' data science courses. It is a data science program after all so I would expect to see this. It is the same for NYU's or Columbia's Data Science programs. I would not say they are as basic as the users above said, as in bootcamp level, but here are a few of the core courses in MIDS for you to check yourself:

1. Machine Learning: https://kylebradbury.github.io/ids705_sp2020/syllabus.html

2. Statistical Modeling: https://ids-702-f20.github.io/Course-Website/

3. Practical Data Science: https://www.practicaldatascience.org/html/index.html This is an intro to programming course similar to 523L in the core Statistics MS at Duke.

 

The second year of the program has 6 electives of your choice, so you actually get to choose 50% of the classes you will take across departments. I personally have chosen electives in the Mathematics, Computer Science, ECE and Statistics departments and I have not experienced any problems getting access into them. During Add-Drop period I added and dropped tons of subjects and I never had a single professor tell me I was not allowed to take their subject. I don't know of anyone at MIDS that had any issues taking the course of their choice  either as long as courses were not closed to students of all other department and as long as students met prerequisites (although, even then, in most cases professors are open to waive those requirements). In fact some MIDS students have gone off to do unusual electives like Real Analysis or Topology. 

 

The program places really well on both Internship and Full-Time opportunities. In fact it places much better than other programs like the Statistics MS. The first cohort secured offers as data scientists or similar in Facebook (2 or 3 students), Google, McKinsey, Bain, Uber, Blackrock... and it was a relatively small cohort of 36 people. I know firsthand this year won't be much different, most people have already secured really good positions despite covid. And no, it is not only the students that had strong backgrounds before coming to the program. You can see for yourself in Linkedin.

 

The program does not have issues receiving applicants. This year they received around 960 applications, for which they plan to extend 30 offers ( with a yield of 50%, that means 15 students), partly because of deferrals from the past year, but also because they really want to keep class sizes small. Acceptance rates are quite similar to what you see on other top Masters at Duke, hovering around 10%.

 

Also, I don't know anything about professors not wanting to teach at MIDS or about the statistics department not wanting anything to do with MIDS. As far as I am concerned, this is actually false with professors in the Statistics department reserving seats for MIDS students in subjects like Bayesian Statistics.

Edited March 1, 2021 by MIDS_Student

[dontoverfit]

Just chiming in here since this topic is somewhat personal to me...I hope the program improves! Duke has a lt of faculty in other departments doing excellent data science research. However, MIDS is still going through growing pains, and I think there are other much better options. This doesn't take away from some really high quality students in the program! And in the interest of getting a diverse collection of opinions, I encourage other students of the program to give their own thoughts! Perhaps others had a better experience.

With that being said, I think it is interesting that someone dug up a year old thread that had "MIDS" in the title...I guess it demonstrates that faculty are aware of damage from negative reviews on GradCafe! ?