Part 5/10:
Central to unsupervised learning is the concept of density modeling. The goal is essentially to ascertain the underlying distribution from which the given data is drawn. Traditional models, including auto-regressive models, utilize a chain rule approach to decompose input into manageable units to predict the next piece of data based upon prior observations.
These models exhibit advantages in various domains—such as generating sequences in natural language processing and coherent images in computer vision. By operating on smaller pieces of high-dimensional data, auto-regressive models tackle some of the challenges posed by the curse of dimensionality inherent in machine learning tasks.
Applications of Unsupervised Learning
- Feature Learning: