Presentation
Have you ever built a machine learning model that performed extremely well on your training and test data, only to see it fail when applied in a new setting? This is a common challenge in machine learning. Models often learn patterns that work well in the data they were trained on but do not persist when the environment changes.
For example, a model trained to predict income from U.S. census data using individuals from some states may not perform equally well in other states. Similar challenges arise when predicting traffic accident severity across regions or forecasting stock price movements across different years. The central question is how to build models that continue to work when conditions change.
In this talk, Camilla will discuss why many predictive models fail to generalize across environments, and how we instead can focus on patterns that remain stable when the world changes. She will introduce Constrained Random Forest, a modification of the random forest algorithm designed to prioritize such stable predictive relationships, and illustrate how this can help build models that continue to work even after they leave their training environments.
Speaker
Camilla Lingj?rde is a DSTrain Postdoctoral Fellow at the Department of Mathematics, UiO. She completed her PhD in Biostatistics at the University of Cambridge in 2023. Her research focuses on statistical and computational methods for analyzing complex biological and biomedical data.
Program
11:30?– Doors open and lunch is served
12:00?– "When machine learning leaves its comfort zone" by Camilla Lingj?rde (Postdoctoral Fellow, Department of Mathematics)
This event is open for all students, PhD candidates, postdocs, and everyone else who is interested in the topic. No registration needed.
About the seminar series
Once a month, dScience will invite you to join us for lunch and professional talks?at the Science Library. In addition to these, we will serve lunch in our lounge in Kristine Bonnevies house every Thursday. Due to limited space (40 people), this will be first come, first served.?See how to find us here.
Our lounge can also be booked by?PhDs and Postdocs on a regular basis, whether it is for a meeting or just to hang out – we have fresh coffee all day long!