A Topological Regularizer for Classifiers via Persistent Homology
This work addresses the need for structure-aware regularization in supervised learning, offering a novel approach to improve classifier simplicity without sacrificing flexibility, though it is incremental in applying topological methods to regularization.
The paper tackles the problem of simplifying the classification boundary by introducing a topological regularizer based on persistent homology, which controls spurious topological structures and is demonstrated to be effective on synthetic and real-world datasets.
Regularization plays a crucial role in supervised learning. Most existing methods enforce a global regularization in a structure agnostic manner. In this paper, we initiate a new direction and propose to enforce the structural simplicity of the classification boundary by regularizing over its topological complexity. In particular, our measurement of topological complexity incorporates the importance of topological features (e.g., connected components, handles, and so on) in a meaningful manner, and provides a direct control over spurious topological structures. We incorporate the new measurement as a topological penalty in training classifiers. We also pro- pose an efficient algorithm to compute the gradient of such penalty. Our method pro- vides a novel way to topologically simplify the global structure of the model, without having to sacrifice too much of the flexibility of the model. We demonstrate the effectiveness of our new topological regularizer on a range of synthetic and real-world datasets.