neu250 - Computational Neuroscience - Statistical Learning (Complete module description)

1 SWS Lecture (VL)
Total workload 36 h: 14 h contact / 22 h individual revision of lecture contents, test preparation, and application to portfolio tasks

1 SWS Seminar (SE)
Total workload 36 h: 14 h contact / 22 h individual reading and test preparation

3 SWS Supervised exercise
Total workload 108 h: 42 h contact/ 66 h individual work on portfolio tasks (programming and interpretation of simulation or analysis results)

attendance in pre-meeting

Upon successful completion of this course, students

• have refined their programming skills (in Matlab) in order to efficiently analyze large-scale experimental data
• are able to implement a processing chain of prefiltering, statistical analysis and results visualization
• have acquired an understanding of the theoretical underpinnings of the most common statistical analysis methods and basic machine learning principles
• have practised using existing toolbox functions for complex analysis tasks
• know how to implement new analysis algorithms in software from a given mathematical formulation
• can interpret analysis results in a neuroscientific context
• have applied these techniques to both single channel and multi-channel neurophysiological data

++ Neurosci. knowlg.
+ Scient. literature
+ Social skills
++ Interdiscipl. knowlg.
++ Maths/Stats/Progr.
+ Data present./disc.
+ Scientific English

• data preprocessing, e.g., artifact detection and rejection, filtering, z-scoring, epoching
• data handling for high-volume data in Matlab
• introduction to relevant analysis toolbox software
• theory of multi-dimensional statistical analysis approaches, such as multi-dimensional linear
• regression, principal component analysis, independent component analysis, logistic regression,
• gradient-based optimization
• practical implementation from mathematical formulation to software code, debugging and unit testing
• postprocessing and results visualization
• consolidation during hands-on computer-based exercises (in Matlab)
• introduction to selected specialized analysis approaches during the seminar

Wallisch et al.: MATLAB for Neuroscientists, 2nd Ed. Academic Press. More text books will be suggested prior to the course. Scientific articles: Copies of scientific articles for the seminar will be provided prior to the course

Recommended in combination with neu240 Computational Neuroscience - Introduction
Shared course components with (cannot be credited twice): psy220 Human Computer Interaction

Course in the first half of the semester Students without Matlab experience should take the optional Matlab course (1. week) of Computational Neuroscience - Introduction

during the course

Portfolio, consisting of daily short tests, programming exercises and short reports