What is the difference between Coursera's data science track (via Johns Hopkins) and...

[Quora Feeds]

Ricardo Vladimiro

Hi and thank you for the A2A.

I didn't do any of Udacity's courses so I have no base to compare quality. Assuming you are mostly interested in content, from reading Udacity's course description and from my experience with Coursera's, the differences I see are:

• Udacity assumes prior knowledge in statistical inference. One of Coursera's courses is of statistical inference. I would say that both assume knowledge of descriptive statistics. Udacity states it, Coursera doesn't if I recall correctly.
• Udacity assumes prior programming knowledge, especially Python. One of Coursera's course is R programming, the only language you will need for the specialisation.
• Coursera is, in my opinion, more academic. It distinguishes data science from big data and focuses on the science part of data science. For instance, it worries with reproducible research and the creation of data products that a data scientist can share and discuss. While that is very useful on a corporate setting, it is (again in my humble opinion) a bigger concern in research and academia context.
• Udacity approach is, in my opinion, more corporate. It exposes the student to things like MapReduce, MongoDB while not forgetting my dearest friend R. It has a visualisation part which is very important for professional. To give you some context, the knowledge I got from reproducible research is very useful to discuss internally at my work with other analysts. We can go through the data, the process and the mind set, apply new data sets and discuss it easily. However, for stakeholders a story telling visualisation is much more important than the ability to reproduce the research.

I hope this helps!

See Questions On Quora

Continue reading...

 

[Quora Feeds]

Allan Reyes


Disclosure: I wrote this when I was still a student/graduate. Since then, I realized that I believed in Udacity's mission so strongly that I applied for a job there--and got in! Everything below this line is the unedited original.

Some context: I've taken Udacity's Data Analyst Nanodegree (DAND) and completed three courses in the Coursera/JHU Data Science Specialization, including Practical Machine Learning (PML). These are, of course, my opinions: I'm partial to Udacity's DAND program.

Udacity's DAND offers a very comprehensive track that is very project-oriented, with coverage on both breadth and depth: Python, scikit-learn, machine learning, R, exploratory data analysis, MongoDB, and visualization libraries like D3.js and dimple.js. JHU's data science track only uses R, but still covers the full gamut of a data analyst's "toolbox." Other than the PML course, I did take the first two: The Data Scientist's Toolbox and R Programming. I was actually disappointed in them, having finished both on the same day that I started. Nonetheless, I think it's more fitting to compare the experience between PML and the DAND machine learning course and project.

The PML course was "more breadth than depth." It was great to get exposure to various machine learning algorithms in the lectures, as well as the theoretical and academic background. However, like the first two courses in the specialization, I did not find the project very challenging. I did not really get a sense of learning or accomplishment after completing the project. The project involved applying a machine learning algorithm to classify physical exercises based on sensor data. Here's a link to my report: Practical Machine Learning Project. It was easy to get 100% prediction accuracy, and it took me only a few hours to complete.

In contrast, the DAND's machine learning course and project were extremely challenging! The project was "Identifying Fraud from Enron Emails and Financial Data": allanbreyes/udacity-data-science. The project was extremely open-ended, and the dataset was both real and challenging. It was difficult to get even 30% prediction precision, and I think that's more representative of real-world data that an analyst would have to hack through in industry. This project easily took 20-40 hours to complete. For me, this was the best way to learn--getting my hands dirty and delving into a challenging problem to apply theory.

I think that describes the core difference between the two programs. JHU's data science track is more academic, whereas Udacity's DAND is more industry- and project-focused. You'll leave JHU with a university certificate, and you'll leave the DAND with a portfolio of projects. (And, career services support to help you land a data analyst job.)

To be clear, I think they're both great programs that are worth doing. If you include the other introductory statistics courses offered at Udacity outside the DAND, I think both programs are beginner-friendly and cover the "full stack" of data science. However, I must admit that my experience with the DAND was more formative. I think a good combo for me was to take the full DAND and use JHU's data science courses to supplement any topics that I wanted further study on.



See Questions On Quora

Continue reading...