I use this site to serve training content to students in my class:
Build Machine Learning Applications with Linux, Python, and Spark
- Using a scenario-based, outside-in development methodology, you will build and assemble modules built from Linux, Python, and Spark into a Machine Learning Application.
- You start by finding observations data on the web which is information rich (financial or sports data are good choices).
- You will use Linux to pull new observations into a data store (CSV, Postgres, or HDFS perhaps) each minute (or hour or day).
- You will use Python (Pandas, NumPy, psycopg2, SFrame, PySpark) to transform observations into taining data and test data.
- You will use Python APIs of respected Machine Learning libraries to learn from data (scikit-learn, Theano, and TensorFlow).
- From your Machine Learning models you will predict past observations and then gauge accuracy and effectiveness of your models.
- You will use Python Data Visualization technology to show model behavior to your end-users: Matplotlib, Bokeh
- You will use Python web technology to serve visualizations (and API data) to your end-users: Django, Flask
- You will use cloud technology to present predictions from your Machine Learning Application to end-users and investors: Amazon EC2, Heroku
- You will use Linux and Python to monitor your Machine Learning Application to maximize its uptime and performance: urllib, BeautifulSoup, Selenium