What are the best distance learning masters programs (in any country) in Machine Learning and NLP?

Question

Randy Crawford · Answer

Stanford and Columbia are at the top of my list, since you are lectured to and tested by their full time profs.  Maybe Illinois?

A cost effective MS in CS?  If you take a job in the financial industry, ANY MS program will be cost effective.  If you work pro bono, none of them will be.  Any distance program won't be cheap.  All are designed to be profit centers.

Chintan Donda · Answer

For Learning,

If U want to Learn along with Recognition after the completion of course,
U can opt for COURSERA courses on Machine Learning & NLP.
They will also give U certification after the completion of ur courses.
And now-a-days almost each IT industry give EQUAL IMPORTANCE to the courses U have completed ONLINE in addition to ur Graduation courses.

Sridhar Mahadevan · Answer

By and large, it is difficult to give meaningful replies to such questions, because the answers are inherently subjective. Rankings mean nothing in practice because programs ebb and flow, and people come in and leave. Particularly in AI and ML, there’s been a net outflow to industry from most academic departments, and so, by and large, many programs have been decimated in the past 10 years. This trend is likely to continue in this superheated hiring climate where industry is prepared to spend lavish sums acquiring AI/ML talent. The surest way to deny yourself a true learning experience is to pay attention to rankings.

But, even more importantly, you should realize that AI is an incredibly diverse field, and fundamental breakthroughs can come from anywhere. Some of the best ideas over the past 10 years have in fact come from outside AI, in fields like optimization (e.g., the framework of proximal algorithms and mirror descent methods), statistics (e.g., regularization methods in ML), and physics (e.g, spin glasses, energy models). So, a department that is likely to have stars in statistics or physics or biology may be just as valuable in studying AI as a program that has strong AI researchers. The most valuable seminar I took a couple of years ago was actually by a professor who taught in the business school, because she was a world class expert in a mathematical framework that generalized optimization in a really cool way. The mathematical framework she was an expert on was originally developed in physics. There she was, teaching in a business school, about ideas that came from physics, and there I sat, as a machine learning person, enthralled by the ideas she was teaching. If you open your mind, you can learn useful ideas about AI from almost every corner of any good university.

The late Caltech physicist Richard Feynman used to often quote Gibbons: “The power of instruction is seldom of much efficacy except in those happy dispositions when it is almost superfluous” (see Preface to his famous Lectures on Physics, 3 volumes). He was a strong believer in self-learning and self-discovery.

In plain English, what this means is that the best students are those who have already learned much of what you are trying to teach them, and need only that last extra percent that you can provide. In all my years of teaching, I’ve felt this captured the dilemma of teaching better than anything else I’ve read.

In short: if you want to learn AI, be your own teacher! There are plenty of online resources: books, tutorials, papers, code. Teach yourself what you need to know. Learn at your pace. Find out what excites you. Self-discovery is often the most exhilarating thing in the world (more than one Nobel Laureate has commented that the moment of discovery can be so exhilarating that it beats almost every human experience, even sex!). Once you experience the true joy of learning on your own, you will find it easy to decide on where you want to go to study AI. You may even end up in a surprising place, like a business school or a biology or statistics department!

George Karabassis · Answer

There are many renowned universities which you can do your PhD in NLP. The most important question is where are you looking to live in. Is it the UK, USA? Elsewhere?

I could go ahead and list some options:

* Carnegie Mellon University
 * Stanford University
 * Edinburgh University
 * University of Washington
 * University of California, Berkeley
But in my opinion, the rankings or the place which you do your PhD doesn’t really matter. You ideally want a good supervisor who is willing to supervise you and give you funding for your project.

My best advice is to look through each university one by one and look into their NLP group faculty and see if any faculty member matches your interest. You shouldn’t choose a faculty member from a top university because it’s simply the “top”.

The rest of my answer will contain bias elements, meaning that you should only read if interested.

Would be difficult to give you an opinion of what’s like being in one of the schools mentioned above. I can only speak of Edinburgh as it’s the institution I’m currently attending.

Edinburgh has the largest AI research facility in the UK, and frankly, has one of the most productive NLP groups worldwide. I have met a number of faculty members working in the NLP group in Edinburgh and definitely are great people to work with. All research is done within the Informatics Forum, next to the Bayes Center and the building definitely feels like being a part of a large family.

Credit: School of Informatics

Good luck with your research!

Julian Cook · Answer

Orig Q: Who is pursuing a masters degree in data science/machine learning online?

I am going to assume, for working purposes, that the implied question is ‘why would you pursue an online Data Science masters?’. Or additionally: ‘Is it worth pursuing one online?’

I see that the other answers from Mohamed Ben Haddou and Julian Torres Santeli are focused on the quality of graduates from these programs, which is a legitimate question.

Some background

1. I’m currently enrolled at Univ Wisconsin, pursuing their MS Data Science program.
2. I had previously done the 10-part Coursera Data Science Specialization, mainly out of interest.
3. I work in NYC
4. I am primarily studying Data Science for the skill set, not for career advancement, though it might enable, for instance, a change to a different field.
The alternatives: Full-time or part-time masters

It would be better, in fact ideal, to stop work for a year and study full time. You get all the advantages of on-campus interaction with professors and classmates. But for me, like most mid-career people, this would be financial suicide. It would be a lot like a 747 with all 4 engines out. This is a common situation for most mid-career people.

Studying part-time and attending evening classes sounds like a great idea, but if you work in NYC, you usually don’t live anywhere near it (I don’t). If I attended evening classes, I would probably be getting home around 11pm at night, 2x per week. Just in time to go to bed. This would get miserable after the first semester, not to mention the slowly rising hostility you would probably receive from your spouse.

So what about studying online?

One thing I learnt from working through the Coursera 10-part ‘specialization’ was that a computationally focused subject, if ‘re-mastered’ for online, is 100% viable and very engaging. The difference appears to be in the staff and how well they prepare and support the material. If you watch an Andrew Ng video online, it becomes obvious that watching him online is far more productive than sitting in a physical classroom with an average professor.

So right here you can see that a well prepared online course can ‘broadcast’ a better quality of education out to a remote and wider audience, but there are some very big blanks still to be filled in. These are coursework, testing and Q&A. The non-degree online MOOCs like Coursera use peer-review, forums and online quizes to accomplish this. It generally works in a non-degree way, but your peers are not paid to expertly assess and advise on your progress or have stimulating off-topic conversations. This is up to you.

The major difference with the university masters degrees, is that intake is restricted - just like normal masters programs. As accredited institutions, they also have to maintain the same academic standards as the equivalent onsite courses. Class sizes are constrained to a level that allows the professor to know and interact with each course participant and give them individual feedback and also allows them to actively participate in forum discussions. My experience is that they take participation very seriously. You probably interact with them more online, than you would physically. Questions get answered at 11pm at night.

The downsides are obviously lack of face-to-face interaction which some institutions like UC Berkeley try to solve with 1–2x yearly onsite weeks, but otherwise you are completely responsible for pacing yourself from week to week.

Big question is; would I hire someone from a online Masters: I would say yes - overall these people had to be quite disciplined to get through the coursework, so they have to be motivated and have the ability to work though problems on their own. In terms of relevant expertise - it would be easy for me, when interviewing, to just look up their courses and quiz them on what they had covered and how much actual research they did personally.

For the non-degree specializations like Udacity or Coursera, I would need to review what each curriculum contained. I know, for instance, if you made it through the capstone project for Coursera, then you are definitely not stupid - but for the others it would require digging into each course and getting to the bottom of what they really did and know. So overall evaluating these candidates is a lot more work than ‘She just graduated from Stanford’.

Overall, if you need the people and you know your subject, I don’t see why it is difficult to evaluate candidates from Data Science programs, from wherever they graduate from.

Santosh Amba Prasad · Answer

In recent years, there are numerous online opportunities that universities have initiated. I will try to answer the question in most practical terms possible.

* Firstly, I’m assuming that you are working full time and your primary goal is to pursue academics to advance your career. (Like me).

* Secondly, what do you want to get out of MSc? Do you want to pursue higher qualifications and get into academics or you want to equip yourself with as much conceptual knowledge as possible and apply this in your professional life.

* Other factor, which is crucial (at least in my case) is financial status - how much you can invest in your academics, keeping in view your current and near future circumstances.

Based on these considerations, I have outlined the options in the order of preference :

a) MSc in Machine learning and Artificial Intelligence from LJMU, Liverpool UK - This program is run by collaboration of Upgrad, India and LJMU, UK and it started in March 2019. Since then, it has picked up a lot of movement. Upgrad is already a well-established and trusted institute in India running Diploma programs with state-of-the-art curriculum in collaboration with premier Indian institutes like IIIT, Bits Pilani, IIT etc. You can get all the details here - Master's in Machine Learning and Artificial Intelligence with IIIT Bangalore | upGrad [ https://www.upgrad.com/masters-in-ml-ai-ljmu-iiitb/ ]. The program is completely online and therefore best suits the working professionals. The rich program curriculum combined with practical knowledge by industry experts is very effective. The post-graduation qualification run by LJMU gives an exposure to full-fledged research. The degrees earned is acceptable at various other UK universities for further research opportunities ( PhD/MRes/MPhil).

b) Master’s Degree from University of Edinburgh - University of Edinburgh is one of the oldest premier British universities. It is in the top 20 universities of the world. The curriculum is not much different than the first option, however you will need to do a lot of paperwork to get the admission. Obviously, if you are not a citizen of Britain, you will have to shell out a lot. I have done PG Certification from this university. I personally feel that the curriculum is more research focused and academic.

c) Options on Coursera - Columbia University/Imperial College - Both the programs are offered on Coursera. At the time, I wrote this answer, Imperial college program is not yet released. Both the universities are top brands, so I guess you will have to dig your pockets deep.

d) Options on edX - University of Texas Austin and Georgia Tech are offering fully online masters programs but they are analytics heavy. You will also have to go through the qualification exams - which I think is a major drawback.

I have contemplated a lot and done a through look out and opted for option (a) – MSc with Upgrad and LJMU. I have learning experience of both Diploma and Master’s program and I personally certify that it has helped me explore further research and publication opportunities and strengthened my career graph.

Noah Gift · Answer

There isn’t one correct answer to this question, it is very situational, just like Machine Learning and Data Science in the real world. A few things to consider are:

Where do you physically live while participating in the program?

Are you going to work full-time or part-time?

What type of job role are you targeting when you finish your degree? Is it business analytics focused, industry focused, or science focused?

I believe there is tremendous value in attending a program where you can work full-time while you are getting a Master’s degree. One, you are still earning income, and two you can apply what you learn to the job you are in, three, your company will often reimburse your fees. I have done this twice, working full-time while getting a Master’s degree and I believe it helped my career tremendously. It isn’t easy, but often worthwhile things take sacrifice.

Mohamed Ben Haddou · Answer

As a CEO of a data science consulting company, my most important Job is hiring data scientists.

I see a lot of candidates following some kind of online learning course or diploma. While this give them some knowledge about the overall process and the details of the algorithms, they usually lack few fundamental skills that are related to hands on practice.

When I hire them, I need to plan few months of training where they work on real problems and are confronted to the reality of data science work.

So my conclusion is, it is certainly a plus, but I will not take it with same face value as a real university certification and much less than hands on experience on real, not on-line projects.

Ross Ramzook · Answer

It really does depend on what you want to achieve by learning more about NLP. Is it that you want to create positive change in the lives of others? Do you want to operate out of an office and be identified as a professional within a niche with a wealth and abundance of knowledge? If so then you probably should pursue a related degree program and most likely an MSC afterwards. It all really depends on what you want to do. A degree isn’t necessary but for knowledge sake, it might be relevant to your goals. Are you more into one and one counseling, coaching or research? Most likely if you are interested in using NLP as a catalyst to enhance the lives of others then coaching or positive psychology may be most relevant to your desired niche.

Justin Fister · Answer

The same way you would start working in NLP with a master’s degree. :)

So, I happen to have a MS degree in CS, but that didn’t get me a job in NLP. What got me working in NLP was taking an interest in NLP and pursuing it. If you are starting from scratch, I would recommend purchasing a popular book on Natural Language Processing or Text Analytics that is code-based, and forcing yourself to work through the entire book. This will give you a firm foundation in the basics and introduce you to areas that may be of interest.

After that, pick a topic in NLP and go for a deep dive. Explore it, code up a solution to a problem in that area. Deploy the code as a website, an app, or a github project. Repeat these steps enough and you will have plenty of experience to get a job in NLP, or perhaps one of your projects will generate enough revenue that you won’t need to look for a job.

Hari Hara Subramanyam Sreenivasan · Answer

There are two different questions you are asking and I will attempt to provide answers for both of them:

1. How can I improve my chances of getting enrolled in a university for a machine learning master’s degree ? (Based on the way that you have worded your question, you are yet to apply for one)
2. What can I do to improve my learning experience and get the most of out of the master’s degree ? (It is beneficial to take steps to ensure you learn as much as possible and be prepared for graduate study)
For the first one, the application strength would depend on a variety of factors but you do have an inherent advantage since you come from a CS background. Your application strength will depend on:

* You undergraduate grades in probability and statistics, linear algebra, calculus (I had a course on introduction to calculus and course on differential equations as part of my undergraduate). Good performance in courses which have the above as prerequisites is also an indicator of mathematical aptitude for machine learning.
 * If you are applying to a university which has the GRE as a requirement, a good score in the quantitative section will be helpful. Doing well in the writing section can give you a small edge but will be looked over as long as your score isn’t abysmal and you do well in quant.
 * A very strong Statement of Purpose. This is an important part of the application and for writing a strong SOP:
1. 
1. I would suggest that you check out a few intro to data science and/or machine learning articles. They can help you understand interesting applications in various fields as well as introduce you to the components of data science (data cleaning, data exploration, data analysis and data visualization)
2. Mention the courses which capture your interest and why they will be useful to your eventual career goals / what you wish to get out of the degree.
3. If you have a lenient word limit, express which electives make sense for your learning and attempt to talk about what you look forward to learning. (I mentioned data visualization and financial electives courses in my data science applications)

Among the above, the grades aren’t in your hands anymore but the other parts are and focusing on them will improve your chances of admission.

The below steps are optional but will be helpful to get the most out of your degree. I would suggest the following:

1. Mathematics: The requirements should have been explored well in a CS degree. Brush up on:
2. 
1. Probability and Statistics. (No need to go above and beyond your UG curriculum but do revise what you have learned will help you hit the ground running) .
2. Linear Algebra. (You should be comfortable with working with matrices and understand their properties)
3. Sets and Calculus.

3. Programming: If you will be working with open source tools, the most popular are R and python. I would recommend Python (I personally use R) as it is better for machine learning (based on popularity and having a community helps when you are stuck or need guidance). Become familiar with the syntax (shouldn’t be a problem since you are a software engineer) and learn to work with the following libraries:
4. 
1. NumPy
2. pandas
3. SciPy

In the theoretical aspect of machine learning, the mathematics will be helpful since the proofs are based on the fact that, as N (the number of data points) gets larger, the algorithm will improve in accuracy (not exactly the right way to put it but this will be discussed in detail).

In the application aspect of machine learning, familiarity with the toolkit will help in maintaining productivity so that you can actually solve the problems and implement solutions rather than focus on learning the tools.

All the above are just suggestions. I didn’t put in too much time to prepare for my graduate course in Data Science and still managed to learn quite a lot but if I had done the above, I would have saved myself a few headaches.

Good luck, and I assure you, it is an interesting field (at least I think so).

Ram Naresh Jha · Answer

Here are some online universities that offer master’s degrees in machine learning and artificial intelligence:

IIT Kanpur

IIT Kanpur: Offers an e-Masters in Artificial Intelligence and Machine Learning.

Johns Hopkins University

Johns Hopkins University: Offers an online master’s degree in Artificial Intelligence.

Georgia Tech

Georgia Institute of Technology: Offers a Master of Science in Computer Science with a specialization in Machine Learning.

Columbia University

Columbia University: Offers online programs in Machine Learning.

Duke University

Duke University: Offers online programs in Machine Learning.

Stevens Institute of Technology

Stevens Institute of Technology: Offers online programs in Machine Learning.

Drexel University

Drexel University: Offers online programs in Machine Learning3.

Colorado State University–Global Campus

Colorado State University Global: Offers online programs in Machine Learning.

University of Michigan–Dearborn

University of Michigan Dearborn: Offers online programs in Machine Learning.

Northeastern University

Northeastern University: Offers online programs in Machine Learning.

Please note that the availability of these programs can change, and it’s always a good idea to check the university’s official website for the most accurate and up-to-date information. Also, consider factors such as course content, faculty, cost, and the support services available to online students when choosing a program. Happy learning! 😊

Pierre Andurand · Answer

To me, the best option is:

1) Columbia’s MSc in Computer science/Machine Learning track through CVN

Computer Science Master's Degree - Machine Learning [ https://cvn.columbia.edu/program/columbia-university-computer-science-masters-degree-machine-learning-masters-science ]

Also to consider:

2) Imperial College’s MSc in Artificial Intelligence which will start in 2020 on Coursera

3) Stanford:

Computer Science MS Degree [ https://online.stanford.edu/programs/computer-science-ms-degree ]

For Data Science

1. UC Berkeley Master Earn Your Master’s in Data Science Online [ https://datascience.berkeley.edu/ ]
2. Johns Hopkins Data Science [ https://ep.jhu.edu/programs-and-courses/programs/data-science#quickset-program_pages_content_4 ]
Good luck!

Arko Neogy · Answer

There are several colleges that have a strong reputation in AI at the Master’s and PhD level. At an undergrad level you are expected to study a diverse mix of subjects so I wouldn’t know.

Here are some of the most well known ones:

1. Stanford (Andrew Ng, Fei Fei Lee, Andrej Karpathy)
2. Berkeley (Pieter Abeel, Michael Jordan)
3. MIT (CSAIL group)
4. CMU (Robotics lab)
5. U Toronto (Hinton)
6. NYU (Yann le Cun)
7. U Montreal (Bengio, Goodfellow)
8. U Washington (Carlos, Emily)
Some people like Goodfellow, Andrej are no longer in academia.

However, I would like to highlight that good research does not ONLY take place where the famous people are.

Zubin John · Answer

GPA's are pretty much what decide how you get in. So here's a walk-down on the procedure (might vary slightly from university to university). The first phase is an academic review where they look at your GPA and GRE scores and decide if you clear a certain cut-off. In the second phase (departmental review) the faculty/ certain PhD students go through your SOP and recommendations and pick the students who interest them. Generally the departmental committee has no idea of your GPA and scores, and their decision is purely on the basis of your writing and recommendations.

That being said if you have a decent GPA (3.0) / (~7.0) then you should probably look into universities that are ranked below thirty. This does not mean they are any less in terms of academic excellence, just that in the past years they have not as many publications in journals as the top 30.

Machine Learning R...

That being said if you have a decent GPA (3.0) / (~7.0) then you should probably look into universities that are ranked below thirty. This does not mean they are any less in terms of academic excellence, just that in the past years they have not as many publications in journals as the top 30.

Machine Learning Research Specific:
Michigan State University
Boston University
U of California - Irvine
U of California - Santa Cruz
Pennsylvania State University
City University of New York
U of Pittsburgh

Natural Language Processing:
University of Minnesota
Boston University
U of Texas - Dallas
University of Rochester
Oregon Health & Science University
University of Utah
City University of New York

Christine · Answer

Here's how you can approach choosing your research direction within cyberbullying detection using NLP for your Master's in NLP:

Consider your supervisor's suggestion:

* Benefits: Your supervisor is likely an expert in the field and can offer valuable guidance and resources throughout your research. Cyberbullying detection is a relevant and impactful area within NLP, with ongoing research and potential for real-world application.
 * Exploration: Discuss your interests within cyberbullying detection with your supervisor. Are there specific aspects (e.g., sarcasm detection, emotional analysis) that particularly appeal to you?
Evaluate your own interests:

* NLP Subfields: Within NLP, are there specific areas you find fascinating (e.g., machine translation, question answering)? Can you find a way to integrate these interests with cyberbullying detection?
 * Research Focus: Do you prefer a more technical approach focused on model development, or a social science angle exploring the impact of cyberbullying detection systems?
Research the Landscape:

* Current Research: Review recent research papers on cyberbullying detection using NLP. What are the common techniques used? Are there any gaps or under-explored areas that interest you?
 * Emerging Techniques: Explore the latest advancements in NLP, like transformers or large language models. Can these be applied to improve cyberbullying detection?
Discuss Options with your Supervisor:

* Present your findings: Summarize your exploration of your supervisor's suggestion and your own interests.
 * Brainstorm collaboratively: Discuss potential research questions that combine your interests with your supervisor's expertise and address gaps in existing research.
Here's a framework to guide your decision:

1. Relevance to NLP: Ensure your research question leverages and contributes to the field of Natural Language Processing.
2. Feasibility: Consider the resources available (data, computational power) and your timeline for completing your Master's research.
3. Originality: Strive for a research question that addresses a gap in existing knowledge or proposes a novel approach.
4. Personal Interest: Choose a topic that you find genuinely interesting and motivating, as this will sustain your focus throughout your research journey.
Remember, your Master's research is an opportunity to explore your interests within NLP and make a valuable contribution to the field of cyberbullying detection. By thoughtfully considering your supervisor's suggestion, your own interests, and the research landscape, you can confidently choose a direction that sets you up for success.

Sushil Sukhwani · Answer

Pursuing a master’s in machine learning from a foreign country can prove to be a good decision. Choosing the best programme to pursue in machine learning can be according to individual preferences and academic interests. Here are some programmes that you can consider pursuing:

* Master 	of Science in Machine Learning (MSML) from Stanford University, USA
 * Master 	of Science in Computer Science (MSCS), Massachusetts Institute of 	Technology (MIT), USA
 * Master 	of Science in Electrical Engineering and Computer Science (MEng) 	from Carnegie Mellon University, USA
 * Master 	of Information and Data Science (MIDS), University of California, 	Berkeley,USA
 * MSc 	in Machine Learning from ETH Zurich, Switzerland
 * Master 	of Science in Data Science from the University of Oxford, UK
 * Master 	of Science in Applied Computing (MscAC) from the University of 	Toronto, USA

Sandeep Singh · Answer

Georgia Tech’s Online MS Course is best as:

* Cheapest
 * Best quality as from top 10 Computer schools in US
 * Flexible with schedule.
 * Amazing community.
Thanks

Aima Aizaz · Answer

Best master program in machine learning? There is no answer to your question, as everyone has a unique career goal. It depends on your career choice (Data scientist, ML engineer, or BI developer), what your goals are, what you want to achieve, etc. There are several universities and platforms offering master's degrees.

1. Columbia University (Computer Science Master's Degree - Machine Learning)
2. Duke University (Data Analytics and Machine Learning)
3. Harvard University (Master in Data Science)
4. Simplilearn
5. Edureka

Petr Baudis · Answer

Learn neural networks for text processing.

Neural networks are becoming increasingly popular for a wide range of NLP tasks, from sentiment analysis to entity/relation extraction to speech recognition. And entering that field has never been easier - with public availability of pretrained word embeddings (word2vec and Glove), user friendly deep learning libraries like Keras and plenty of data like the Wikipedia yours for taking.

Neural networks as typically used for processing written text take characters or words in turn, use vector space embeddings to convert them to numbers, and produce an abstract numerical representation of the block of text, which captures the aspects of its meaning that you care for. They can process the text either as variable-width n-grams (convolutional neural networks), by iteratively updating the meaning representation word by word (recurrent neural networks), or by building up the meaning from the parse tree (recursive neural networks). Even bag-of-words approaches can get very powerful when slanted as mean-pooling neural networks.

Neural networks are not the panacea in all areas of NLP, at least not yet. But they are getting more versatile by the day and other weak points like dataset size constraints are getting better. The good part is that once you learn to think in terms of neural networks, any new NLP problem you meet will be easy to prototype in days. Of course, there’s always a long journey from a prototype to business-quality tool, but that’s the case for all approaches. And in some sense, the neural networks are getting insanely powerful and appear eerily smart.

Final advice? Stay driven by practical problems! Look at competitions or academic benchmarks, try to figure out something fun. Predict rating of Wikipedia articles or user’s intent in chat, tag product listings on Amazon, solve History SAT exams, auto-caption images in Quora answers, align audiobooks with text or use them to learn to read text in an actor’s voice, summarize news stories, the ideas are endless.