- The problem with supervised learning: Creating labelled data is challenging, time-consuming, and often requires expertise (e.g., healthcare). Consequently, labelled databases are typically small.
- The Unsupervised Opportunity: Unlabelled data is available in abundance.
- The Goal: Learn about structure of data. Since most random pixel configurations do not look like real-world images, data is highly structured.
- Representation Learning: The objective is to learn a compressed representation (code) that preserves useful information about the input.