Some posters may be thinking that by a certain point, many CS majors are ready to go deep in a way that only a school with graduate program resources can support. NLP for example was characterized above as a “vertical” specialty. Seems to me, it could just as well be approached as a “horizontal” (integrating CS, math/stat, language(s), and linguistics). Many LACs could support study in all those areas (with the possible exception of linguistics) sufficiently for a strong undergraduate interdisciplinary program. An example of what they probably can’t do is train a student deeply in machine learning approaches for speech or OCR processing. If that’s the kind of experience you think you’ll want (and be ready for by the end of your 3rd year or so), go to a tech institute or a big comprehensive RU. Or get it in grad school.
CS programs at highly selective LAC’s are requiring 11-12 CS courses for a major. Unlike many uni’s that allow one to get credit for HS Calc, most highly selective LAC’s don’t play that game - no credit for 5’s on AP exams, although it will allow one to start at a higher level introductory Calc class if they are able to demonstrate the same proficiency on the school’s placement exam.
^ As that goes, even calc I on the college level will not count toward the math major at many LACs. Some degree of advanced placement for certain majors will be expected.
Let’s not paint all highly selective LACs with a broad brush.
Williams and Amherst (both notable highly ranked LACs) only require 8 CS classes. I could probably find even more with similarly light CS curricula if I looked into it.
There are many benefits from attending a LAC, a comprehensive CS curriculum is usually not one of them. Let’s not forget that some of you are comparing highly selective LACs with normally selective RUs, the disparities will be even greater when you start considering less selective LACs.
@yikesyikesyikes, and other highly selective LAC’s require more and consistently place students in the top CS graduate programs in the country - Swarthmore is an example of such a program.
I think @OHMomof2 framed it well:
I would add that specialty courses at RU’s are usually smaller and are taught for professors with subject matter expertise. LACs are hamstrung in their ability to offer deep expertise in multiple topics due to faculty limitations. (see multiple posts on the CS faculty size and course availability at Amhurst).
I disagree with @merc81 who wrote:
Perhaps that argument can be made, but that would be in a different thread. When it comes to CS there are specific metrics that can be compared and contrasted. That’s what were, doing here.
Concerning comment by @tk21769 on CS specialties:
I know a student who is taking the following 2 courses in the upcoming sophomore fall quarter:
Machine Learning
Artificial Intelligence: Principles and Techniques
This is not an unusual course sequence. I think people may be using breadth and depth differently. In terms of CS, LAC resources cannot provide a broad choice of specialty courses (CS breadth). Additionally, aside from basic CS, they cannot really provide depth in a given specialty (CS depth).
Great debate though.
Swarthmore has 7 professors for 40 majors and 10 minors - I can’t imagine a better learning environment.
@Rivet2000 wrote
As far as I can tell, Stanford and Amherst have exactly the same number of courses in Artificial Intelligence:
https://www.amherst.edu/academiclife/departments/courses/1718F/COSC/COSC-241-1718F
https://web.stanford.edu/class/cs221/
Swarthmore offers courses in machine learning, AI, NLP, Bioinformatics, and Adaptive Robotics. Here’s the website for a professor who teaches machine learning and computational biology.
https://www.cs.swarthmore.edu/~soni/
Davidson offers courses in AI, Bioinformatics, and Machine Learning. Here’s a website for a professor who teaches AI: http://academics.davidson.edu/math/ramanujan/
Middlebury offers courses in AI, Bioinformatics & Biometrics, Machine Learning,
Here’s a list of publications by a professor who teaches a 25-student course in machine learning:
http://www.cs.middlebury.edu/~schar/pubs.html
Carleton offers courses in AI, NLP, and Computational Biology.
Pomona offers courses in AI, Neural Networks, Computer Vision and Image Procession, Robotics, Machine Learning, NLP and Computational Semantics.
Oberlin offers courses in AI, Machine Learning, and NLP.
I wouldn’t expect to see multiple courses or multiple faculty covering each of these areas at many LACs. However, I’m seeing professors with degrees from top universities, more or less extensive publication histories, and maybe some good research opportunities for students. Now, what I’m listing are some of the most selective schools.
Shall we do a side to side comparison of Stanford AI track vs Amherst AI track? I don’t think Amherst has one, so probably an unfair question. AI is a field that includes many topics so you need to search deeper that just on “AI” classes. Also, simply consider faculty. Looking at Stanford AI lab alone reveals:
Faculty and Research Scientists: 15
Affiliated Faculty: 12
Faculty Emeritus: 5
Looking at the larger CS faculty:
Regular Faculty: 60
Lecturers: 12
Faculty Emeritus: 16
Courtesy Faculty: 24
Adjunct Faculty: 9
Visiting Faculty: 5
@Rivet2000 wrote
Idk, @Rivet2000, you sort of keep moving the goal posts a little. If your point is that Stanford has more faculty that’s fine, if you intend to stay there long enough to enroll in every course they teach. By my estimation, even if you take a full load of, say, eight semester courses a year, you would have to be there ~sixteen years in order to accomplish that feat.
If your point is that it’s impossible for an LAC student to pursue a comp sci topic in depth, you haven’t shown that yet.
Dig a little deeper and I think you find that the depth of courses on certain topics is greater. Comparing Amherst to Stanford is probably a little unfair though. I’m sure students at both schools are receiving outstanding educations.
Those two AI courses are not even comparable. The Stanford one is more rigorious and has heavier prereqs. Please read through the syllabi. A true introduction to AI requires a good CS knowledge base. Amherst lets you do it before learning a good amount of data structures and algorithms and before knowledge of any discrete mathematics/structures. That would be unfathomable to people at any half-decent RU with familiarity in CS.
@yikesyikesyikes wrote
Well, I guess, if at first you can’t succeed, try, try and try again. First, the inference is that there are no courses at all in AI or ML at top LACs like Amherst, Davidson, Swarthmore, Middlebury, Carleton, Pomona or Oberlin. When that is shown not to be true, the next tactic is to attack the rigor of the courses offered. Typical Big U bombast.
It does not surprise me at all that Stanford has prerequisite requirements for an introductory course in AI. It’s a sexy course in an enormously popular department on a university campus that worships the ground Hewlett and Packard walked on. How else are they going to weed out the hundreds of students who just want to be able to say, “I took AI at Stanford.”?
FWIW, there are no shortage of algorithm courses at top flight LACs:
Wesleyan
https://iasext.wesleyan.edu/regprod/!wesmaps_page.html?stuid=&facid=NONE&crse=003338&term=1181
https://iasext.wesleyan.edu/regprod/!wesmaps_page.html?stuid=&facid=NONE&crse=003338&term=1181
Williams
https://catalog.williams.edu/catalog.php?strm=1181&subj=CSCI&cn=134&sctn=D1&crsid=020808
https://catalog.williams.edu/catalog.php?strm=1183&subj=CSCI&cn=256&sctn=01&crsid=010806
https://catalog.williams.edu/catalog.php?strm=1181&subj=CSCI&cn=353&sctn=T1&crsid=020807
Swarthmore
https://www.cs.swarthmore.edu/~adanner/cs41/f06/index.php
Carleton
https://apps.carleton.edu/campus/registrar/schedule/enroll/?term=18SP&subject=CS&number=252
Oberlin
http://obiemaps.oberlin.edu/courses/introduction-to-algorithms/
https://presto15.cc.oberlin.edu:8074/pls_live/bwckschd.p_disp_listcrse?term_in=201802&subj_in=CSCI&crse_in=365&crn_in=%
Please do not stoop down to using a strawman argument. I never said that AI courses are not offered. I also never said LACs do not offer courses in data structures and algorithms.
You cannot possibly have a legit AI course without knowledge of data structures + algorithms and discrete structures/mathematics. This is the case in the course YOU linked to. Data structures and algorithms may be offered, but if students are not required to take it before AI, are they really learning AI at a level rigorious enough for CS majors? The answer is an objective no.
Doctorates Awarded to Graduates of Liberal Arts Colleges, FY 2014
Top 25 Liberal Arts Colleges
http://www.swarthmore.edu/sites/default/files/assets/documents/institutional-research/BaccOrigins14.pdf
Math and CS are clubbed together, so difficult to say how many got their Doctorates in CS
For the top 100 institutions by discipline (page 26 for Math and CS)
http://www.swarthmore.edu/sites/default/files/assets/documents/institutional-research/Doct%20Rates%20Rankings%20by%20Broad%20Disc%20Field-Summary%20to%202015.pdf
Guess Stanford kids don’t have too much of an incentive to go for Doctorates 
^^Agree on the doctorates statement. By that time most Stanford grads have already launched their startups and are too busy talking with all the VCs 
LOL… I work for a VC firm in Palo Alto !
Percentage of Doctorates from various institutions is a red herring for disciplines like Eng and CS as some of the smartest folks start working directly after their undergrad (or even before completion!)
Not every LAC is like Amherst.
Not every research university is like Stanford.
And very few students get admitted to either Amherst or Stanford anyway.
Why do people here persist in arguing about LACs versus research universities as if they were all the same within each grouping?
What I find frustrating are the people who think that by going to a big school that offers more CS classes, it means you’ll come out of it a better programmer than if you’d gone to a small school. If you take the same number of CS classes at a big or small school, the only advantage the big school confers is that you may be able to tailor a program more to your liking. Or you may not. It depends on how flexible the particular CS program in a school is.
But you won’t necessarily come out of a big, research university with more CS knowledge. That will depend on the quality of the teaching, and how much time you spend in front of the computer trying to get your programs to work.