A Review of the Udacity Deep Learning Foundations Nanodegree
James Wrubel

Great review, James Wrubel! I’m one of your fellow January cohort graduates.. Here’s my review of the course:

The Deep Learning Nanodegree Foundation course is a great introduction to the rapidly-evolving field of modern AI (in a big-picture sense), and in particular to ‘deep learning’ fundamentals and several important types of neural networks. In this ambitious offering, Sebastian Thrun and the course creators aim to democratize the AI skill set and spark interest in further AI-related education & careers.

The course content consists of a collection of video lectures, text sections, links to external references, quizzes and projects. In addition to the steady guidance from lead instructor Mat, the highlights are some of the guest instructors and all the projects, which present a reasonable challenge but also “training wheels” in various forms. Contrary to the current misleading title, which as of May 2017 still reads “Siraj Raval’s Deep Learning”, the video presentations from Siraj basically supplement the actual course content. The originally-advertised component of live sessions with Siraj was subsequently relegated to “optional” status or as a suggested activity for advanced students who enjoy coding challenges. I believe the live sessions are being discontinued for future cohorts, and will be replaced with recorded videos from Siraj.

Given that I went through the inaugural cohort of this course, there were of course “growing pains” with respect to the course format, content, delivery and teaching quality. These were offset to varying degrees by the fact that the lead instructors and program coordinators were quite receptive to student feedback via Slack and Waffle.io, and generally did their best to respond to student concerns. Having the Slack community and Udacity discussion forums proved to be a great source of support for learning, brainstorming, venting and overcoming challenges.

A note on the prerequisite knowledge and time commitment: I would consider intermediate Python knowledge, a Linear Algebra refresher and Matrix Multiplication to be high-priority prerequisites. Knowledge of calculus & partial derivatives would also be helpful, but it’s possible to get by as a beginner without fully understanding the advanced math portions of the lectures. As far as time commitment, the course FAQs have now been updated to state average 8–12 hours weekly, which is more realistic than their original estimate of 3–5 hours a week. I would budget more like 15–20 hours a week for newcomers.

One clap, two clap, three clap, forty?

By clapping more or less, you can signal to us which stories really stand out.