Unsupervised Learning
Table of Contents
1. Basics
The major point of differentitiation initiates from how the relevant dataset is structured.
1.1. Dataset
- a collection of unlabelled examples (only feature vectors)
1.2. Objective
- train a model using the dataset such that it transforms feature vectors into practical representations
- these representations are task dependent
- Clustering : a categorical cluster label
- Dimensionality Reduction : a semantically useful condensed vector
- Outlier Detection : an indicator as to how different a feature vector is from the usual instance in the dataset