Universal probability-free prediction
This work addresses the need for assumption-free prediction methods in machine learning, offering a foundational approach that could impact various domains, though it appears incremental by building on existing theories.
The paper tackles the problem of constructing universal prediction systems without statistical assumptions, based on Popper's falsifiability and Kolmogorov complexity, and shows that under the IID assumption, these systems dominate conformal prediction within usual accuracy.
We construct universal prediction systems in the spirit of Popper's falsifiability and Kolmogorov complexity and randomness. These prediction systems do not depend on any statistical assumptions (but under the IID assumption they dominate, to within the usual accuracy, conformal prediction). Our constructions give rise to a theory of algorithmic complexity and randomness of time containing analogues of several notions and results of the classical theory of Kolmogorov complexity and randomness.