Department Lunch Talk! Kevin Rosenfield

On February 10th @2pm, Kevin Rosenfield will give a talk titled “Artificial intuition: Can machine learning techniques help us understand the evolutionary origins of concealed and advertised ovulation?” via Zoom.

Hypotheses regarding the evolutionary origins of behavioral and soft-tissue traits are notoriously difficult to test, and those invoking natural selection normally often depend on untestable assumptions, such as the continuity of traits’ functions over time. While the absence of fossilized remains limits our ability to study such traits directly, modeling techniques make it possible to test many components of evolutionary hypotheses. For instance, agent-based models simulate the interactions between individuals and their environments given a set of predetermined traits in order to study the more general patterns that emerge from these interactions. One limitation of such models is that they are parameterized by humans, who inevitably introduce their own biases. Revolutionary new machine learning techniques, such and deep and reinforcement learning, have been shown to surpass human intuition in many domains, allowing us to uncover patterns that have not previously been considered. Using agent-based modeling combined with machine learning, I am developing models to test hypotheses of the evolutionary origins of concealed and advertised ovulation.
Kevin Rosenfield is a current Ph.D. candidate at Penn State University. His dissertation research utilizes agent-based modeling simulations to test hypotheses about the evolution of behavioral and morphological (physical) traits in animals, particularly non-human primates.