Unsupervised learning

x is output

Evaluation is important than loss minimization since the loss is unlikely to directly capture the desired result due to the lack of the label. Therefore, contextualizing and interpreting the inference result is important.

How data is composed

Data Clustering

Dimension Reduction

Representation Learning

Explicit density methods (explicit model of
Likelihood)

Autoregressive Model

Normalizing flow

AutoEncoder

Diffusion Model

Implicit density methods (implicit model of
Likelihood)

Energy-based Model

Using Large Language Models Effectively

What 10 Cutting-Edge Companies Have Learned From Building on Top of LLMs

https://unsupervisedlearning.substack.com/p/using-large-language-models-effectively

Unsupervised learning

x is output

Explicit density methods (explicit model of Likelihood)

Implicit density methods (implicit model of Likelihood)

Recommendations

Explicit density methods (explicit model of
Likelihood)

Implicit density methods (implicit model of
Likelihood)