Columbia Video Network is the online arm of Columbia Engineering, and it has been running since 1986, which makes it one of the oldest distance programs in the country. The original model was literally videotapes mailed to corporate partners within driving distance of Manhattan. Today it is streaming, but the underlying idea has not changed: CVN students watch the same lectures as the students sitting in the room, submit the same homework, and sit the same exams. The MS in Computer Science requires 30 points, which is ten courses at the 4000 level or above, with at least six points at the 6000 level. CVN students must choose one of the department's tracks, and the Machine Learning track is the one most readers here will care about.
It asks for two fundamental courses, for example COMS 4771 Machine Learning or 4776 Neural Networks and Deep Learning or 4252 Computational Learning Theory, then two secondary electives with at least one at the 6000 level, and the balance in general electives. There is also an NLP track and a Vision, Graphics, Interaction and Robotics track if you would rather specialise differently. The on-campus professional development requirement does not apply to CVN students. The cost is where I have to be blunt. CVN lists tuition at $2,700 per point, plus a $395 CVN fee for every course and a one-time $105 transcript fee.
Columbia's own estimate for the full MS is $85,055. There is also a $1,000 deposit when you accept and an $85 application fee. That is Ivy League on-campus pricing for an online degree, and it is roughly four times UCLA's online engineering master's, around eight times UT Austin's MSAI and more than ten times Georgia Tech's OMSCS. Admissions are less brutal than Stanford but not casual. Columbia asks for a GPA of at least 3.3, three recommendation letters, transcripts, a resume and a personal statement.
The GRE is optional. Decisions are rolling, you can begin in fall, spring or summer, and admitted students have up to a year to start, which is genuinely helpful if you are waiting on a tuition benefit to kick in. What I like is that this is not a watered-down version of the degree. The diploma does not say online, the courses are the real graduate courses, and the ML track sits on top of one of the stronger CS departments in the country. If the Columbia name is what gets your resume read in your city or industry, this delivers it. What I like less is the delivery.
Lecture capture is honest but dated. Some courses are filmed live during the semester and you can interact with the current instructor and TAs. Others are pre-taped from an earlier year, and students have complained for years that those feel disconnected from the live class, the current homework and the people grading it. The practical advice from alumni is to favour live-taped sections where you can.
Videos are also only available while you are enrolled in the course, so you cannot go back to them later. My honest take is that CVN is a good degree sold at a price that only makes sense with someone else paying. If your employer covers it, take it and enjoy the name. If not, OMSCS and UT Austin exist, and the gap in what you will learn is far smaller than the gap in what you will pay.