Let (X,dX) and (Y,dY) two nonempty compact Polish spaces, µ 2M +1 (X), 2M +1 (Y) two probability measures on these spaces and c: X Y! R+ a nonnegative and continuous function. As X and Y are compact, r(µ,) is tight, then Prokhorov's theorem applies and the closure of r(µ,) is sequentially compact. Let us now show that r(µ,) is closed. Indeed, Let ( n)n 0 a sequence of r(µ,) converging towards . In addition as ( n)n 0 live in the simplex r, we can also extract a sub-sequence, such that n! 2 r.

The matching principles behind optimal transport (OT) play an increasingly important role in machine learning, a trend which can be observed when OT is used to disambiguate datasets in applications (e.g.

We show the full derivation of Equation 6 as follows. Recall that X0i is the concatenation of Xi and Tl. The model variation used here in TransEvolve-fullFF. Thus, on the limiting case, we get E[Ul(Ul)>] = 1I where I is the d-dimensional identity matrix. This way, Ul2 dd approximates a rotation matrix as we choose σ = O(d).

The Transformer and its variants have been proven to be efficient sequence learners in many different domains. Despite their staggering success, a critical issue has been the enormous number of parameters that must be trained (ranging from 107 to 1011) along with the quadratic complexity of dot-product attention. In this work, we investigate the problem of approximating the two central components of the Transformer -- multi-head self-attention and point-wise feed-forward transformation, with reduced parameter space and computational complexity. We build upon recent developments in analyzing deep neural networks as numerical solvers of ordinary differential equations. Taking advantage of an analogy between Transformer stages and the evolution of a dynamical system of multiple interacting particles, we formulate a temporal evolution scheme, TransEvolve, to bypass costly dot-product attention over multiple stacked layers. We perform exhaustive experiments with TransEvolve on well-known encoder-decoder as well as encoder-only tasks. We observe that the degree of approximation (or inversely, the degree of parameter reduction) has different effects on the performance, depending on the task. While in the encoder-decoder regime, TransEvolvedelivers performances comparable to the original Transformer, in encoder-only tasks it consistently outperforms Transformer along with several subsequent variants.

Tree Search (TS) is crucial to some of the most influential successes in reinforcement learning. Here, we tackle two major challenges with TS that limit its usability: distribution shift and scalability. We first discover and analyze a counter-intuitive phenomenon: action selection through TS and a pre-trained value function often leads to lower performance compared to the original pre-trained agent, even when having access to the exact state and reward in future steps. We show this is due to a distribution shift to areas where value estimates are highly inaccurate and analyze this effect using Extreme Value theory. To overcome this problem, we introduce a novel off-policy correction term that accounts for the mismatch between the pre-trained value and its corresponding TS policy by penalizing under-sampled trajectories.

Meta-learning approaches enable machine learning systems to adapt to new tasks given few examples by leveraging knowledge from related tasks. However, a large number of meta-training tasks are still required for generalization to unseen tasks during meta-testing, which introduces a critical bottleneck for real-world problems that come with only few tasks, due to various reasons including the difficulty and cost of constructing tasks. Recently, several task augmentation methods have been proposed to tackle this issue using domain-specific knowledge to design augmentation techniques to densify the meta-training task distribution.