Best (US) Graduate Schools for Programming Languages
So, you want to know what the best graduate schools for programming languages are. Well, probably no one has the “correct” answer, but let’s try to estimate it.
First, let’s try to understand what a “good” graduate school is. In my personal opinion, two parameters define the caliber of a graduate school: the volume of the research it produces, and the quality of this research[1]. Well, the research volume is commonly measured by the number of papers a school publishes. The quality of the research is harder to estimate, and I discuss this below.
How can we measure the quality of a single research paper? I hope most people will agree that “quality research” means “influential research”—research that influences, or impacts the work of, many researchers. Hence, it would be safe to say that quality, or influential, papers are cited more frequently than papers of lower quality.
With this observation in mind, we have the means to estimate the caliber of a graduate school: a school with high caliber will have a lot of papers that are cited frequently. So, the metric for how good a school is, is the weighted number of papers the school has published as follows:
where:
- ShoolPublication is the set of papers published by the school.
- Wp is the weight of a paper p, or the number of papers that cite paper p.
Frequently Asked Questions
- How do you count the number of papers that cite a particular paper?
Well, you got me. I do not have the time to count the papers that cite a particular paper. Instead, I weight a paper according to the conference it appeared in[2]. The main assumption is that quality papers usually appear in quality conferences. To determine the quality of a conference I used the list at http://citeseer.nj.nec.com/impact.html which ranks conferences according their impact. For my purposes, I consider conference with high impact to be high quality conference too.
In summary:
Let IC be the impact given at http://citeseer.nj.nec.com/impact.html for conference C. If a paper, p, appears in C then Wp=IC.
- Some papers come from two (or more) schools. For example,
Parametric Shape Analysis via
3-Valued Logic; Mooly Sagiv
(Tel-Aviv University); Thomas Reps (University of Wisconsin); Reinhard Wilhelm (University of Saarland), POPL 1999.
Do you count the paper in the caliber of all three schools?
Yes and no. Such papers will be
counted for the SchoolCaliber of all three
schools. However, since
In summary,
Let
SÎ{1,2,...}
be the ‘position’ of the school in the author list, then 
.
- What about research labs? Do you rank them too?
Currently, no.
- Does the affiliation of an author changes when the author moves to another school?
No. With respect to my analysis, an author's affiliation is determined at the time a paper is published. Hence, WP is fixed forever.
The Results
- Conferences in the database:
1. PLDI1996, PLDI1997, PLDI1998, PLDI1999, PLDI2000, PLDI2001, PLDI2002, PLDI2003, PLDI2004
2. POPL1994, POPL1996, POPL1997, POPL1999, POPL2000, POPL2001, POPL2002, POPL2003, POPL2004.
- Total number of papers in the database: 520.
- Best 30 schools:
|
1 |
Carnegie
Mellon |
77.88 |
|
2 |
|
65.20 |
|
3 |
MIT |
53.39 |
|
4 |
|
50.18 |
|
5 |
|
32.45 |
|
6 |
Rice |
27.31 |
|
7 |
|
25.90 |
|
8 |
Cornell |
25.56 |
|
9 |
|
23.43 |
|
10 |
Stanford |
23.29 |
|
11 |
|
23.14 |
|
12 |
|
19.58 |
|
13 |
|
17.47 |
|
14 |
McGill |
16.31 |
|
15 |
Tel-Aviv |
14.57 |
|
16 |
Northeastern |
14.43 |
|
17 |
|
13.72 |
|
18 |
|
13.67 |
|
19 |
|
13.56 |
|
20 |
|
13.25 |
|
21 |
Chalmers |
13.25 |
|
22 |
Harvard |
12.91 |
|
23 |
Yale |
12.32 |
|
24 |
|
11.56 |
|
25 |
|
10.80 |
|
26 |
Purdue |
10.80 |
|
27 |
|
10.68 |
|
28 |
|
10.21 |
|
29 |
Technion |
9.79 |
|
30 |
Oregon
Graduate Institute |
9.79 |