Unbalanced Sobolev Descent
This addresses the challenge of distribution alignment in machine learning and computational biology, particularly for tasks like matching developmental stages of cells, but it is incremental as it builds on existing particle descent algorithms with a novel discrepancy measure.
The paper tackles the problem of transporting high-dimensional source distributions to target distributions with potentially different masses by introducing Unbalanced Sobolev Descent (USD), a particle descent algorithm that combines advection and reaction steps, and demonstrates faster convergence than previous methods on synthetic examples and in single-cell RNA sequencing applications.
We introduce Unbalanced Sobolev Descent (USD), a particle descent algorithm for transporting a high dimensional source distribution to a target distribution that does not necessarily have the same mass. We define the Sobolev-Fisher discrepancy between distributions and show that it relates to advection-reaction transport equations and the Wasserstein-Fisher-Rao metric between distributions. USD transports particles along gradient flows of the witness function of the Sobolev-Fisher discrepancy (advection step) and reweighs the mass of particles with respect to this witness function (reaction step). The reaction step can be thought of as a birth-death process of the particles with rate of growth proportional to the witness function. When the Sobolev-Fisher witness function is estimated in a Reproducing Kernel Hilbert Space (RKHS), under mild assumptions we show that USD converges asymptotically (in the limit of infinite particles) to the target distribution in the Maximum Mean Discrepancy (MMD) sense. We then give two methods to estimate the Sobolev-Fisher witness with neural networks, resulting in two Neural USD algorithms. The first one implements the reaction step with mirror descent on the weights, while the second implements it through a birth-death process of particles. We show on synthetic examples that USD transports distributions with or without conservation of mass faster than previous particle descent algorithms, and finally demonstrate its use for molecular biology analyses where our method is naturally suited to match developmental stages of populations of differentiating cells based on their single-cell RNA sequencing profile. Code is available at https://github.com/ibm/usd .