Syllabus#

Title: S Project Seminar Biopsychologie - Computational Ethology
ID: 112726
Credits: 5
Semester: WS 21/22 Ruhr-University Bochum

Description#

What is behavior, and how can we measure it? Experimental Psychologists operationalize behavior as task relevant interactions with stimuli and rewards, such as frequency of key pecking, reaction times, and the number of errors in specific tasks, while biologists classify behavior in qualitative clusters and analyze time spent on grooming behavior, or the frequency and intensity of aggressive interactions. New advances in computer vision and machine learning have changed the way we measure spatiotemporal dynamics of animal movement, but do we really understand Animal Behavior? In this seminar we will work with video data of human and non-human animals in different settings and apply cutting edge machine learning techniques to extract spatial and temporal data to describe the continuous stream of behavior.

Expectations#

Programming skills are not necessary, but technical affinity and basic computer skills will be advantageous. The course language may be either English or German, depending on students’ background. The seminar will be graded by (1) active participation in classroom interactions and (2) a final poster presentation. Students will have to read some of the provided literature prior to the respective paper-discussions. Wherever possible, the content of the seminar will be tailored to prospective bachelor dissertation projects. After this seminar students will have learned to use Python for data analysis, as well as some state-of-the-art machine learning techniques for computational ethology such as DeepLabCut (Mathis et al., 2018), Anipose (Karashchuk et al., 2020) and VAME (Luxem et al., 2020).

Contents#

  1. Introduction to Computational Ethology

    • From Tinbergen to Deep Learning

    • Methods and goals

  2. Why Tracking is not Behavior

    • Differences in location, kinematics and poses

    • Multi-animal settings

  3. Levels of Analysis and Quantification of Behavior

    • Structure, spatial relation, and consequence

    • Measuring latencies, frequencies, intensities, or duration?

  4. Classification of Animal Behavior

    • Ethograms and Expert-Annotations

    • Supervised and Unsupervised machine learning

  5. The Problems of Spacetime

    • Spatiotemporal Dynamics

    • Sampling Methods and Dimensionality Reduction

  6. Measuring in Multidimensional Space

    • Goal oriented research and lost Dimensions

    • 3D Imaging, Triangulation and 2D lifting