Linear Algebra Module 3.1 .

Multioutput Regression and Classification using pseudo inverse

1. Basic linear algebra concepts are revised and examples of computational experiments to perform for verification of concepts are shown.
2. How to modify coefficient estimation in multivariate linear regression problem to multiclass classification task is shown next with examples
3. Linear and non-linear Kernel method is introduced for classification and solved using pseudo inverse. The method introduced do not require constrained optimization theory as in classical theory for support vector machines and kernel methods. Just assume unknown coefficient vector is in rowspace of data matrix or kernel matrix.
4. Explicit mapping of data to higher dimension (random kitchen sink algorithm) followed by regression for classification is done next with coding examples.
5. Finally examples of creating own data sets for classification and clustering task is shown.