In this project, I make a revision about the theory of the Spectral Clustering technique, which receives a graph representing the data as input and outputs a clusterization based on Normalized Cuts in the graph (this notion is explained in the project).
The project goes through concepts from the original paper and subsequent descriptions provided to it, expected awareness as "how can I represent my data as a graph?" or "how can I compute the Laplacian matrix for large graphs?", a bonus theorem to estimate a lower bound of $k$, i.e., the number of clusters, and an application example for object detection in images.