Member-only story
Machine Learning Strategy — Part 1
I am currently Taking Deeplearning.ai Deep Learning Specialization which I highly recommend by the way for people who are looking for a great place to start deep learning. The Specialization is offered on Coursera and I have just finished their third Course : Structuring Machine Learning Projects.
The course explains how to be systematic when it come to thinking about Deep Learning projects and gives you an array of tools that’ll help you to make the right decisions and move forward with your machine learning project.
Most of the ideas of this course are not included in university deep learning classes. Thus, I am writing this blog post to sum it up for you.
Click here for Part 2 — (I am still on it, will be there soon)
Table of Content of Part 1
- Why ML Strategy
- Orthogonalization
- Single number evaluation metric
- Satisfying and Optimizing metric
- Train/dev/test distributions
- Size of the dev and test sets
- When to change dev/test sets and metrics
- Why human-level performance?
- Avoidable bias
- Understanding human-level performance
- Surpassing human-level performance
- Improving your model performance
