Naturalistic animal behavior is complex. Traditionally, there have been two general approaches to dealing with this complexity. One approach, common in psychology, is to simplify an animal’s environment, or its movements, in order to make precise measurements. Another approach, taken by ethologists, is to study complex naturalistic behaviors directly. In many cases this choice has forced researchers to give up on quantitative rigor. Recent breakthroughs in camera technology and computational techniques open up the possibility of merging these approaches. We can now describe naturalistic behavior quantitatively.
Students will be expected to engage with the material, and discuss with their peers and the instructor during class. Homework is in the form of reading papers that will be discussed the following class (~2 hours/week), and in learning the background concepts necessary to understand and discuss the papers (~2 hours/week). Projects ideas will be proposed in writing ~2/3 way through the course (citing the relevant literature), and project results will be presented to the class. Projects will be assessed based on how they demonstrate the student’s mastery of the relevant course material, creativity, and on presentation quality.
Background
Traditional approaches to ethology and neuroscience
Basics of data analysis, machine learning, deep neural nets
Basics of optics, computer vision, camera design
Quantifying movement
Image filtering and morphological operations
Marker based and marker-less pose estimation
2.5 dimensional imaging (RGB-D)
Semantic segmentation
Tracking collective animal behavior
Describing behavior
Eigenworms and the dynamical systems view
Mouse behavioral syllables and Markov models
Drosophila behavioral space and nonlinear dimensionality reduction
Supervised learning approaches (e.g. boosted decision trees)
Drosophila social behavior and generalized linear models
Linking behavior to neural activity
Traditional brain-behavior correlations: hippocampal cell types
Imaging neural activity in freely moving C. elegans
Selective neural activation-behavior screens in Drosophila
Cephalopod skin patterning dynamics
Mouse basal ganglia and motor control
Collective behavior
Boids model
Experimental manipulation of animal collectives
Fish schooling and deep attention networks
Noise-induced schooling
The material builds on basic knowledge of linear algebra, machine learning, neuroscience, and behavioral ecology. A background in any of these topics isn’t required if a student is willing to learn the relevant concepts as they arise, in preparation for discussing papers in class.