Roadmap for beginners

• Machine learning in 6 months
• Machine learning in 3 months

Theory

Learning materials

Online courses

• Coursera, Machine learning specialization, University of Washington This the most basic course. It covers all the basics and has a few programming tasks. Math knowledge needed is covered in the course materials. Technology: Jupyter / IPython.
• Coursera, Andrew Ng, Machine learining, Stanford University This course is recommended for all beginners. It skims over math except for matrix operations which are covered in the course materials. There are some programming assignments, but rather an afterthought. Technology: Octave, If you want to go deeper, try Andrew Ng’s Stanford lectures
• Coursera, Neural networks for machine learning, University of Toronto This is a tough one. One needs to understand partial derivatives quite well and algebra (dot product, etc. It is not covered in the course) to effectively follow it. Programming part is really tiny and based on Octave. You probably need pen&paper rather than your computer.
• Coursera, Probabilistic Graphical Models, Stanford University I recommend doing this course before the neural networks one as this clarifies Bayes models which are used all over the place in the second part o NN. Navie Bayes is covered as well.
• Coursera, Deep-Learning

Lectures

• Carnegie Mellon University, 10-601 Machine learning
• Carnegie Mellon University, 10-702/36-702, Statistical machine learning

Practice

• Awesome machine learning

Data sets for training and experimentation

• UCI machine learning repository
• Kaggle
• KDNuggets
• Are there any alternatives for Kaggle?

Software aids

• Scikit: great docs and used all over the place
• Anaconda: Python distribution packed with all you need for ML (SciPy, etc). Especially usefull on Windows, where building C-deps is usually a struggle.
• Spark MLLib
• TensorFlow: Low level stuff GPU computation aid from Google
• Keras: Due to TensorFlow being kind of low level
• RStudio: Statistical computing suite. Probably a better fit for ML than Octave, but octave is simple and covers all beginners’ needs.
• Fast artificial neural network library: This is C, but has binding for virtually all programming environments
• Google cloud machine learning platform

Other

• 7 steps to mastering machine learning with Python
• Best practices from Google
• Infographic: A Beginner’s Guide to Machine Learning Algorithms