The characteristic of unsupervised learning is that, unlike in supervised learning or reinforcement learning, the system receives no direct feedback: it is only presented with inputs but not with desired outputs, a reward signal, etc.
The goal of the learning system is instead inherent to it and can usually be described as some kind of loss function that is to be minimized. Unsupervised learning tasks include but are not limited to, e.g.:
- Clustering;
- Anomaly detection;
- Vector quantization;
- Dimensionality reduction;
- …