Course 3: Structuring Machine Learning Projects Week 1 Notes

[mbhushan]

Reducing avoidable bias and variance:

Problems where machines do better than humans.

Surpassing human level performance.

Summary of bias/variance:

Error Analysis Example:

Human Level Error as proxy for Bayes Error:

Available bias and variance:

Why compare to human level performance:

Comparing to Human level performance:

Cat classification - dev/test set:

Orthogonalization of Cat pictures:

Classification Errors:

Splitting data strategies:

Dev/Test set guidelines:

Cat classification test/dev set:

Satisficing and Optimizing Metrics:

Using a single number evaluation metrics:

Chain of assumptions in ML:

Orthogonalization - Motivating Example:

Motivating Example for structuring machine learning projects:

[mbhushan]

Motivating Example for structuring machine learning projects: