Jan 04, 2018
I recently finished the Udacity Machine Learning Nanodegree and wanted to share the following thoughts on my experience.
If you haven't read my Review of Data Science Programs I would do so first, to get a general explanation of the other types of machine learning curriculum.
The Machine Learning Nanodegree is formatted in the previous style of Udacity nanodegrees where you're charged 200 dollars a month until all seven projects are completed. The projects themselves range in complexity, starting from a basic linear regression, to clustering, to more challenging ones such as a reinforcement learning implementation, and a CNN implementation.
The final portion is a capstone project where you are allowed to perform a machine learning project on any dataset of your choosing.
Overall I found the Machine Learning Nanodegree to be pretty comprehensive. Nearly all machine learning topics were covered. Expect this Nanodegree to take quite a bit of time, it's challenging from both a code perspective and from a mathematical understanding perspective. The knocks I have against the program are the level of hand holding, sometimes a bit too much code is implemented for you, and if you're lazy you can get away with just filling in the blanks without understanding the whole picture. However if you're diligent and spend time understanding all the code, even the portions implemented for you, you'll come out with a comprehensive understanding of how to implement Machine Learning projects.
If you're lightly interested in Machine Learning I would recommend Andrew Ng's Coursera course over this one. But if you'd like to dive a bit deeper and have money to spare you can't go wrong with this one as well.