Small satellites enable many important applications for both economic and scientific purposes. Many of these applications are inherently data-centric and rely on large amounts of high-resolution satellite imagery to be delivered in a timely manner. However, communicating this data to Earth is challenging due to intermittent connectivity, high packet losses, low data rates, and similar issues. Therefore, efficient onboard prioritization and data processing are essential for future satellite missions. Machine learning methods, such as deep neural networks, are very suitable for such prioritization, as they are already used extensively for satellite imagery processing and they can be deployed onboard of satellites. However, updating them to support new classification requirements when the satellite is already in orbit is difficult, as often multiple passes are required to complete model transmission due to the communication challenges. To cope with this issue, we propose a progressive transmission mechanism for model updates, which leverages vector quantization and arithmetic coding. Our mechanism allows to achieve high accuracies even with partially updated models. Evaluation results show that our mechanism significantly outperforms other less optimized transmission schemes.

Generalized Additive Models (GAMs) have recently experienced a resurgence in popularity due to their interpretability, which arises from expressing the target value as a sum of non-linear transformations of the features. Despite the current enthusiasm for GAMs, their susceptibility to concurvity — i.e., (possibly non-linear) dependencies between the features — has hitherto been largely overlooked. Here, we demonstrate how concurvity can severly impair the interpretability of GAMs and propose a remedy: a conceptually simple, yet effective regularizer which penalizes pairwise correlations of the non-linearly transformed feature variables. This procedure is applicable to any differentiable additive model, such as Neural Additive Models or NeuralProphet, and enhances interpretability by eliminating ambiguities due to self-canceling feature contributions. We validate the effectiveness of our regularizer in experiments on synthetic as well as real-world datasets for time-series and tabular data. Our experiments show that concurvity in GAMs can be reduced without significantly compromising prediction quality, improving interpretability and reducing variance in the feature importances.