Quantifying uncertainty is an essential part of predictive modeling, especially in the context of high-stakes decision-making. While classification output includes data uncertainty by design in the form of class probabilities, the regression task generally aims only to predict the expected value of the target variable. Probabilistic extensions often assume parametric distributions around the expected value, optimizing the likelihood over the resulting explicit densities. However, using parametric distributions can limit practical applicability, making it difficult for models to capture skewed, multi-modal, or otherwise complex distributions. In this paper, we propose optimizing a novel nondeterministic neural network regression architecture for loss functions derived from a sample-based approximation of the continuous ranked probability score (CRPS), enabling a truly distribution-free approach by learning to sample from the target's aleatoric distribution, rather than predicting explicit densities. Our approach allows the model to learn well-calibrated, arbitrary uni- and multivariate output distributions. We evaluate the method on a variety of synthetic and real-world tasks, including uni- and multivariate problems, function inverse approximation, and standard regression uncertainty benchmarks. Finally, we make all experiment code publicly available.

In automated radio network control, understanding the effect of different factors on network performance is crucial. Although there are machine learning (ML) solutions that can reliably anticipate network performance expressed as key performance indicator (KPI) values, these models are typically black-box or provide only partial explanations. Most approaches are based on flat data structures and cannot exploit the graph nature of the data, being unable to quantify neighbors’ impact. In this paper, we propose a new graph neural network-based model called Ensemble Graph Attention Network (Ensemble GAT) for network KPI prediction. We show that the proposed model results in better or comparable KPI prediction performance to the state-of-the-art while also carrying information about links between neighboring cells. In addition to model creation, we investigate how explainable AI solutions can be used to provide root-cause explanations for network KPI degradation. To generate feature attributions, we introduce an adapted version of GraphLime that works with ensemble models. In addition, we propose a new technique called Neighbor Perturbation to identify the neighboring cells that have the most significant impact on KPI prediction. We demonstrate the effectiveness of these models on both synthetic and real-world datasets.

In this work, we examine Asymmetric Shapley Values (ASV), a variant of the popular SHAP additive local explanation method. ASV proposes a way to improve model explanations incorporating known causal relations between variables, and is also considered as a way to test for unfair discrimination in model predictions. Unexplored in previous literature, relaxing symmetry in Shapley values can have counter-intuitive consequences for model explanation. To better understand the method, we first show how local contributions correspond to global contributions of variance reduction. Using variance, we demonstrate multiple cases where ASV yields counter-intuitive attributions, arguably producing incorrect results for root-cause analysis. Second, we identify generalized additive models (GAM) as a restricted class for which ASV exhibits desirable properties. We support our arguments by proving multiple theoretical results about the method. Finally, we demonstrate the use of asymmetric attributions on multiple real-world datasets, comparing the results with and without restricted model families using gradient boosting and deep learning models.