Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces Value-Driven Transport (VDT), a generative modeling framework based on a stochastic control formulation of measure transport, solved via a linear program. The dual variables of this LP correspond to the optimal value function, which directly encodes the optimal control policy for transporting samples from a source to a target distribution. Experiments demonstrate that VDT policies yield straight transport paths, are simulation-efficient, robust, and can be easily extended with techniques like conditional generation and classifier-free guidance.
Forget flows and diffusions: Value-Driven Transport offers a surprisingly simple and efficient route to generative modeling via optimal control and linear programming.
We propose a new framework for generative modeling based on a discrete-time stochastic control formulation of measure transport. Adapting classic results from control theory, we formulate our problem as a linear program whose dual variables correspond to the \emph{optimal value function} of the control problem, which directly encodes the optimal control policy. Exploiting this LP formulation, we develop an efficient simulation-free primal-dual algorithm for computing approximately optimal value functions and the associated \emph{value-driven transport} (VDT) policies which approximate the true optimal policy. We show that well-trained VDT policies enjoy numerous favorable properties in comparison with other state-of-the-art methods based on flows, diffusions, or Schrödinger bridges: they lead to straight transport paths which can be simulated quickly and robustly, and can be enhanced in all the same ways as diffusion and flow-based models (e.g., conditional generation, classifier-free guidance, unpaired data-to-data translation are all easy to incorporate). We evaluate our methodology in a range of experiments, with results that indicate strong performance and good potential for scalability.