Search

Landslides refer to occurrences of massive ground movements due to geological (and meteorological) factors, and can have disastrous impact on property, economy, and even lead to loss of life. The advances of remote sensing provide accurate and continuous terrain monitoring, enabling the study and analysis of land deformation which, in turn, can be used for possible landslides forecast. Prior studies either rely on independent observations for displacement prediction or model static land characteristics without considering the subtle interactions between different locations and the dynamic changes of the surface conditions. We present DyLand -- Dynamic Manifold Learning with Normalizing Flows for Land deformation prediction -- a novel framework for learning dynamic structures of terrain surface and improving the performance of land deformation prediction. DyLand models the spatial connections of InSAR measurements and estimates conditional distributions of deformations on the terrain manifold with a novel normalizing flow-based method. Instead of modeling the stable terrains, it incorporates surface permutations and captures the innate dynamics of the land surface while allowing for tractable likelihood estimates on the manifold. Our extensive evaluations on curated InSAR datasets from continuous monitoring of slopes prone to landslides show that DyLand outperforms existing bechmarking models.

We introduce a slope-aware graph neural network (SA-GNN) to leverage continuously monitored data and predict the land displacement. Unlike general GNNs tackling tasks in the plain graphs, our method is capable of generalizing 3D spatial knowledge from InSAR point clouds. Specifically, we structure of the land surface, while preserving the spatial correlations among adjacent points. The point cloud can then be efficiently converted to a near-neighbor graph where general GNN methods can be applied to predict the displacement of the slope surface. We conducted experiments on real-world datasets and the results demonstrate that SA-GNN outperforms existing 3D CNN and point GNN methods.

Program semantic properties such as class names, method names, and variable names and types play an important role in software development and maintenance. Method names are of particular importance because they provide the cornerstone of abstraction for developers to communicate with each other for various purposes (e.g., code review and program comprehension). Existing method name prediction approaches often represent code as lexical tokens or syntactical AST (abstract syntax tree) paths, making them difficult to learn code semantics and hindering their effectiveness in predicting method names. Initial attempts have been made to represent code as execution traces to capture code semantics, but suffer scalability in collecting execution traces. In this paper, we propose a hybrid code representation learning approach, named Meth2Seq , to encode a method as a sequence of distributed vectors. Meth2Seq represents a method as (1) a bag of paths on the program dependence graph, (2) a sequence of typed intermediate representation statements and (3) a sentence of natural language comment, to scalably capture code semantics. The learned sequence of vectors of a method is fed to a decoder model to predict method names. Our evaluation with a dataset of 280.5K methods in 67 Java projects has demonstrated that Meth2Seq outperforms the two state-of-the-art code representation learning approaches in F1-score by 92.6% and 36.6%, while also outperforming two state-of-the-art method name prediction approaches in F1-score by 85.6% and 178.1%.