Outline of lecture

Inferring an underlying input-output dependency from input and output examples is called supervised learning. This course focuses on a statistical approach to supervised learning and introduces its basic concepts as well as state-of-the-art techniques.

Purpose of lecture

Inferring an underlying input-output dependency from input and output examples is called supervised learning. This course focuses on a statistical approach to supervised learning and introduces its basic concepts as well as state-of-the-art techniques.

Plan of lecture

1. Introduction
2. Statistical Formulation of Supervised Learning
3. Linear, Kernel, and Non-Linear Models
4. Least-Squares Learning
5. Weighted Least-Squares Learning
6. Regularization Learning
7. Sparse Learning
8. Robust Learning
9. Error Back-Propagation Algorithm
10. Cross-Validation
11. Input-Dependent Estimation of Generalization Error
12. Active Learning
13. Concluding Remarks and Future Prospects

Textbook and reference

None. Handouts are distributed if necessary.

Related and/or prerequisite courses

Probability and Statistics, Pattern Recognition

Evaluation

Small reports related to machine learning and students' projects.

Comments from lecturer

Statistical machine learning is an interdisciplinary subject with a wide range of applicability. Not only learning basic foundations of machine learning, but also applying the learned knowledge to their own research topics is expected.

Supplement

[Office Hours]
Anytime if available.

Pattern Information Processing