Inhalt
zur Navigation
Machine Intelligence I (supervised methods)
General information
- The courses Machine Intelligence I and II can be heard independently.
- Information regarding the tutorials can be found on the ISIS page.
- Lecture and tutorials are held in English, oral exams can be taken in English or in German.
- First day of lecture/tutorials is October 20, 2011.
Topics covered
- Connectionist neurons
- Feedforward multilayer networks
- Learning and generalization
- Radial Basis Functions
- Elements of learning theory
- Support Vector Machines
- Uncertainty and inference
- Bayesian networks
- Bayesian inference and neural networks
Prerequisites
- Mathematical knowledge: analysis, linear algebra, probability calculus and statistics.
- Basic programming skills, preferably Matlab.
Target Audience / Assessment and Grading
| Program | Form of Assessment |
|---|---|
| MSc in Computational Neuroscience | The two courses (I and II) form a single module (12 ECTS). 60% of all assignments & oral exam |
| MSc in Computer Science | Each of the two courses (I or II) can be taken as a separate module (6 ECTS). 60% of all assignments & oral exam |
| Diplom-Informatik | Wahlfach: separate modules possible Graded course credit for the tutorials (25%) and oral exam (75%) |
| Schwerpunkt: separate modules possible Oral block exam, assignments not mandatory | |
| Other study programs | Each of the two courses (I or II) can be taken as a separate module (6 ECTS). 60% of all assignments & oral exam |
Supplemental material
- Table of contents
- Bibliography
- Suggested readings
- MLP gradient descent batch learning
- Proof of the theorem by Robbins and Monro
- Proof of the key theorem of Statistical Learning Theory (SLT)
- Proof of the consistency of ERM
- Number of linearly separable assignments (Proof)
- Proof of an SLT inequality
- Derivation of the 'dual problem' for C-SVM's
- Deviations from the optimal model
- Lagrange parameter and empirical error
- MAP method-result of the integration
- Graphical Models - Slides 1
- Graphical Models - Slides 2
For further information please consult Klaus Obermayer (lecturer, in charge) or Wendelin Böhmer (tutor).