AI Paper Spotlight: Curve Your Enthusiasm: Concurvity Regularization in Differentiable Generalized Additive Models

Come join us for the upcoming session of AI Paper Spotlights - Berlin's AI Research Meet Up, where we'll be delving into the latest advancements in the dynamic field of AI

During this presentation, Alma Lindborg (Machine Learning Researcher at Merantix Momentum), will be discussing the paper titled "Curve Your Enthusiasm: Concurvity Regularization in Differentiable Generalized Additive Models". This research zeroes in on Generalized Additive Models (GAMs), a category of machine learning models highly valued in contexts demanding a high level of transparency in decision-making, such as hiring processes or the medical field. While it's straightforward to visualize how each factor (e.g., age, blood test values) contributes to a prediction (e.g., illness) in these models, Alma demonstrates that these interpretations can be misleading due to concurvity, which refers to mutual dependencies between the features. She will also illustrate that through the regularization of concavity, we can guide the model to establish a more consistent relationship between features and predictions, resulting in a more dependable interpretation.

Register here to secure your spot! You can also participate virtually!

‍

About the speaker:

‍Alma Lindborg is a Machine Learning Researcher at Merantix Momentum) with a background in applied mathematics and cognitive neuroscience. Alma aims to develop ML methods for impactful applications in digital health and life sciences. Her PhD work focused on biophysical signals, and at Merantix Momentum, specializing in ML techniques for sensor data. She has a strong foundation in both applied science and mathematics, with a track record of publications in reputable journals and conferences.