The talk will review the application of kernel based methods to pattern analysis problems. In particular it will emphasize the flexibility and modularity of the resulting algorithms, and their capability to operate on very general types of data, from strings to graphs, from text documents to images. Standard machine learning algorithms like Support Vector Machines will be discussed, as well as statistical methods like Ridge Regression and Correlation Analysis. Finally, the talk will address applications to bioinformatics and text categorization, and connections with Optimization theory.