Coursera offers an excellent machine learning course by Andrew Ng. The course provides comprehensive coverage of topics ranging from linear regression, through neural networks, support vector machines, to unsupervised learning. It also comes with a number of exercises that are a perfect complement to the lectures. Only completing those exercises gives you a much greater mastery of various techniques. In 2015, Google opened Tensorflow machine learning API. This blog rewrites solutions to problems presented in ML lectures, but not the assignments(!), using Tensorflow. Tensorflow makes solutions to the original ML examples much simpler. However, more importantly, if one has understood Andrew Ng’s course, seeing the same problems expressed using Tensorflow provides a better platform for learning and understanding Tensorflow APIs themselves.

Note: Originally, this blog was to be devoted solely to Andrew Ng’s machine learning course. However, recently I decided to add examples of convolutional neural networks, auto-encoders, recurrent neural networks, etc. If you are just interested in material corresponding to Courser’a course, skip ahead to Linear Regression with Multiple Variables in Tensorflow.