Now that we understand some of the tools from TensorFlow that can help us in developing and deploying models at scale, let's try to understand the general rules of thumb when building AI applications.
- Engineering over machine learning: Almost all the solutions to problems start with engineering. It is very important to get the data pipeline right before building any machine learning model.
- Keep it simple: Generally, data scientists have a natural tendency to build the most complex model for the problem. However, it is great to start with a simple, interpretable model—say, a logistic regression model for classification. It helps in discovering and debugging data or engineering pipeline issues better. Only when you are not satisfied with the results of the basic model should you use advanced techniques like deep learning.
- Distributed processing: In the era of big data, you will almost always run into issues where you can't fit the data into RAM. Learning...