Global information is essential for dense prediction problems, whose goal is to compute adiscrete or continuous label for each pixel in the images. Traditional convolutional layers in neural networks, initially designed for image classification, are restrictive in these problems since the filter size limits their receptive fields. In this work, we propose to replace any traditional convolutional layer with an autoregressivemoving-average (ARMA) layer,anovelmodule with an adjustable receptive field controlled by the learnable autoregressive coefficients.

Graph Structure Learning (GSL), a family of data-centric learning approaches, has garnered substantial attention in recent years. The core concept behind GSL is to jointly optimize the graph structure and the corresponding GNN models.

However, the lack of expert-annotated, open-access resources has hindered the analysis of ALS data using advanced deep learning techniques.

Despite their promising performance, the learned knowledge remains implicit in these black-box neural structures, which hinders understanding the importance of input features and how they influencedecisions.