Aggregating Predictors: A Theoretical Perspective
Name of applicant
Mikael Møller Høgsgaard
Title
Soon to graduate PhD-student
Institution
University of Oxford
Amount
DKK 2,713,000
Year
2025
Type of grant
Internationalisation Fellowships
What?
In learning theory, the aim is to define a mathematical link between the amount of training data and the accuracy of a given prediction model. This relationship enables either ensuring a specific prediction accuracy with a given amount of training data or calculating the amount of data needed for a target accuracy. This project seeks to explore these relationships for aggregation of predictors.
Why?
The hope of establishing such relationships between the amount of training data provided to a prediction model and its accuracy is to gain insights into when and how learning occurs and what can be inferred from data.
How?
To establish such relationships, one assumes an underlying learning model or task and then, based on this model, derives mathematical guarantees on the relationship between the amount of training data provided to a prediction model and its achieved accuracy.