We present PyTorch Geometric Temporal a deep learning framework combining state-of-the-art machine learning algorithms for neural spatiotemporal signal processing. The main goal of the library is to make temporal geometric deep learning available for researchers and machine learning practitioners in a unified easyto-use framework. PyTorch Geometric Temporal was created with foundations on existing libraries in the PyTorch eco-system, streamlined neural network layer definitions, temporal snapshot generators for batching, and integrated benchmark datasets. These features are illustrated with a tutorial-like case study. Experiments demonstrate the predictive performance of the models implemented in the library on real world problems such as epidemiological forecasting, ride-hail demand prediction and web-traffic management. Our sensitivity analysis of runtime shows that the framework can potentially operate on web-scale datasets with rich temporal features and spatial structure.

No abstract

Understanding the biases in Deep Neural Networks (DNN) based algorithms is gaining paramount importance due to its increased applications on many real-world problems. A known problem of DNN penalizing the underrepresented population could undermine the efficacy of development projects dependent on data produced using DNN-based models. In spite of this, the problems of biases in DNN for Land Use and Land Cover Classification (LULCC) have not been a subject of many studies. In this study, we explore ways to quantify biases in DNN for land use with an example of identifying school buildings in Colombia from satellite imagery. We implement a DNN-based model by fine-tuning an existing, pre-trained model for school building identification. The model achieved overall 84% accuracy. Then, we used socioeconomic covariates to analyze possible biases in the learned representation. The retrained deep neural network was used to extract visual features (embeddings) from satellite image tiles. The embeddings were clustered into four subtypes of schools, and the accuracy of the neural network model was assessed for each cluster. The distributions of various socioeconomic covariates by clusters were analyzed to identify the links between the model accuracy and the aforementioned covariates. Our results indicate that the model accuracy is lowest (57%) where the characteristics of the landscape are predominantly related to poverty and remoteness, which confirms our original assumption on the heterogeneous performances of Artificial Intelligence (AI) algorithms and their biases. Based on our findings, we identify possible sources of bias and present suggestions on how to prepare a balanced training dataset that would result in less biased AI algorithms. The framework used in our study to better understand biases in DNN models would be useful when Machine Learning (ML) techniques are adopted in lieu of ground-based data collection for international development programs. Because such programs aim to solve issues of social inequality, MLs are only applicable when they are transparent and accountable.

No abstract

The use of stationary, active acoustics provides an effective approach to characterize and monitor temporal variability in the abundance and behavior of pelagic organisms, especially in seasonally ice-covered waters of high latitude marine ecosystems. However, point measurements from stationary echosounders are limited in their spatial coverage. A quantification of the spatial area represented by point measurements (i.e., representative range) is required to ensure effective biological characterization and monitoring. Here, concurrent mobile and stationary active acoustic data collected during summers of 2015 and 2017 are used to assess the representative range of fish and zooplankton density measurements from the Chukchi Ecosystem Observatory located at Hanna Shoal, Northeast Chukchi Sea. Six methods used to calculate representative ranges of backscatter means and variances resulted in representative ranges between approximately 0.3 and 86 km, depending on the year and calculation method. Such relatively large representative ranges reflect the tight bio-physical associations and large characteristic environmental length scales of the NE Chukchi Sea. Between years, up to 10-fold variations in representative ranges were attributed to interannual changes in water mass characteristics and associated species assemblages. Differences of 1–2 orders of magnitude in our calculated ranges among methods are attributed to differences in the rationale and associated assumptions of each approach. The choice of method and resulting representative range depends on monitoring goals: detection of change, mapping of spatial distributions, characterization of spatial variance, or interpolation of temporal variability over space. Our comparison of stationary acoustic to mobile surveys extends the understanding of spatiotemporal variability of marine organism distributions in the NE Chukchi Sea and informs cost-effective design of observing systems to monitor and predict impacts of environmental change.