We propose a gradient-enhanced algorithm for high-dimensional function approximation. The algorithm proceeds in two steps: firstly, we reduce the input dimension by learning the relevant input features from gradient evaluations, and secondly, we regress the function output against the pre-learned features. To ensure theoretical guarantees, we construct the feature map as the first components of a diffeomorphism, which we learn by minimizing an error bound obtained using Poincaré Inequality applied either in the input space or in the feature space. This leads to two different strategies, which we compare both theoretically and numerically and relate to existing methods in the literature. In addition, we propose a dimension augmentation trick to increase the approximation power of feature detection. A generalization to vector-valued functions demonstrate that our methodology directly applies to learning autoencoders. Here, we approximate the identity function over a given dataset by a composition of feature map (encoder) with the regression function (decoder). In practice, we construct the diffeomorphism using coupling flows, a particular class of invertible neural networks. Numerical experiments on various high-dimensional functions show that the proposed algorithm outperforms state-of-the-art competitors, especially with small datasets.

This paper presents a Bayesian model for the multiple-target tracking problem that handles a varying number of splitting and merging targets applied to convective cloud tracking. The model decomposes the tracking solution into events and target states. The events include target births, deaths, splits, and merges. The target states contain both the target positions and attributes. By updating the target attributes and conditioning the events on their updated values, we can include high-level domain knowledge into the system. This strategy improves the tracking accuracy and the computational efficiency since we focus only on likely events for each situation. A two-step multiple-hypothesis tracking algorithm has been developed to estimate the model state. The proposed approach is tested by both simulation and real data for mesoscale convective system tracking.