Learning term structure models

This project develops a new kind of model to better understand how interest rates change over time not just a single rate, but the full range of rates that apply over different time horizons. This structure, often referred to as the term structure of interest rates, affects everything from savings accounts to government policy. Most existing models face a trade-off: they either stick closely to economic principles, respond flexibly to real data, or remain simple to use in practice but not all three at once. This project aims to break that limitation by building a model that meets all three goals: it respects core financial logic, adapts to data without relying on fixed formulas, and remains practical and efficient. One key innovation is the use of a modern tool called path signatures, which allows the model to learn patterns in the data in a structured and reliable way. The result will be a flexible system that can draw on market data, internal patterns, or both to explain how interest rates behave. At present, no open, practical implementation exists that brings these ideas together. This project will develop the necessary theoretical foundations and deliver a working tool for researchers, analysts, and policymakers who need dependable models grounded in both data and financial principles.

No additional funding sources recorded.