Abstract:We demonstrate the relevance of an algorithm called generalized iterative scaling (GIS) or simultaneous multiplicative algebraic reconstruction technique (SMART) and its rescaled block-iterative version (RBI-SMART) in the field of optimal transport (OT). Many OT problems can be tackled through the use of entropic regularization by solving the Schrödinger problem, which is an information projection problem, that is, with respect to the Kullback--Leibler divergence. Here we consider problems that have several affine constraints. It is well-known that cyclic information projections onto the individual affine sets converge to the solution. In practice, however, even these individual projections are not explicitly available in general. In this paper, we exchange them for one GIS iteration. If this is done for every affine set, we obtain RBI-SMART. We provide a convergence proof using an interpretation of these iterations as two-step affine projections in an equivalent problem. This is done in a slightly more general setting than RBI-SMART, since we use a mix of explicitly known information projections and GIS iterations. We proceed to specialize this algorithm to several OT applications. First, we find the measure that minimizes the regularized OT divergence to a given measure under moment constraints. Second and third, the proposed framework yields an algorithm for solving a regularized martingale OT problem, as well as a relaxed version of the barycentric weak OT problem. Finally, we show an approach from the literature for unbalanced OT problems.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is dealing with problems with affine constraints in optimal transport (OT). Specifically, the paper proposes an algorithm named Generalized Iterative Scaling (GIS) or Simultaneous Multiplicative Algebraic Reconstruction Technique (SMART) and its Rescaled Block - Iterative SMART (RBI - SMART), to solve the optimal transport problem with multiple affine constraints. These problems are usually solved by entropy regularization, that is, solving the Schrödinger problem, which is an information projection problem regarding the Kullback - Leibler divergence. However, when it comes to multiple affine constraints, the traditional alternating information projection method may not be directly applicable because these projections cannot be explicitly calculated in general. Therefore, the paper proposes a method of using the GIS or SMART algorithm to replace these projections and proves the convergence of this method.
### Main contributions of the paper:
1. **Generalization of algorithms**: The paper generalizes the GIS and SMART algorithms to the optimal transport problem, especially for dealing with cases with multiple affine constraints. This enables more complex OT problems to be solved, not just marginal constraints.
2. **Convergence proof**: The paper provides a convergence proof of the algorithm. By interpreting the GIS iteration as a two - step affine projection in an equivalent problem, it proves the effectiveness of the algorithm.
3. **Application examples**:
- **Minimizing regularized OT divergence**: Under a given measure, find a measure that satisfies specific moment constraints and minimizes the regularized OT divergence from the given measure.
- **Martingale OT problem**: In financial mathematics, solve the standard OT problem with additional affine constraints, that is, the mean of the target position of each source point should be equal to the source point itself.
- **Weak barycenter OT problem**: Solve the relaxed version of the barycenter OT problem when the cost function depends on the means of the source and target positions.
- **Unbalanced OT problem**: Deal with the unbalanced OT problem with non - exact marginal constraints.
### Formula summary:
- **Kullback - Leibler divergence**:
\[
\text{KL}(p, q)=\begin{cases}
\sum_{i = 1}^M p_i\log\frac{p_i}{q_i}-p_i + q_i&\text{if }0\leq p, q\text{ and }p\ll q\\
\infty&\text{otherwise}
\end{cases}
\]
- **Gibbs kernel**:
\[
K_{ij}=\exp\left(-\frac{c_{ij}}{\varepsilon}\right)
\]
- **GIS iteration**:
\[
p^{(k)}=p^{(k - 1)}\odot\exp\left(A^T\log\left(\frac{b}{A p^{(k - 1)}}\right)\right)
\]
- **Sinkhorn iteration**:
\[
u^{(k)}=\frac{\mu}{K v^{(k - 1)}}, \quad v^{(k)}=v^{(k - 1)}\odot\exp\left(A^T\log\left(\frac{b}{A (v^{(k - 1)}\odot K^T u^{(k - 1)})}\right)\right)
\]
### Conclusion:
The paper provides an effective method to solve the optimal transport problem with multiple affine constraints by introducing the GIS and SMART algorithms. These methods are not only strictly proven theoretically but also demonstrated in practical applications.