Lecture

Technical notes:

1. Re-present input as coefficients of a linear and sparse combination on basises (key input vectors) – encoding process. 
2. On unlabelled data, learning automatically for the basises, also known as features, using neural network. Instead of conventional method to specify the features.
3. Then apply the conventional Machine learning on top of the transformed labelled data for the actual learning task

Remarks:

• Diversify your learning, read more on many topics (unlabelled data) – to implicitly form a way to see (encode)