article-journal

The paper proves that the inverse flow matching problem has a unique solution in one dimension and for Gaussian distributions but the general multidimensional case remains unsolved.

We show that action matching yields optimal transport when using optimal vector fields.

We show that unpaired super-resolution with GANs inherently learns optimal transport maps, but these are biased and do not correctly transform distributions. Recent neural OT methods provide unbiased alternatives and reveal a direct connection to regularized GANs.

We study the impact of emotions on an eSports team performance. By analysing the real data collect at an eSports tournament, we find that there is a strong correlation among the performance of the team, communication between the players, and emotional sentiment of communication.

We develop the General Hedging algorithm for online expert weight allocation under delayed feedback. The algorithm is based on the exponential reweighing of experts’ losses. It extends classical Hedge and Fixed Share algorithms.

We extend the setting of the online prediction with expert advice to function-valued forecasts. As an application of our main result, we prove that various loss functions used for probabilistic forecasting are mixable (exp-concave).

We propose an algorithm for aggregating experts’ predictions which does not require a prior knowledge of the upper bound on the losses. The algorithm is based on the exponential reweighing of experts’ losses.