RESUMEN
In this study, we investigate the communication networks of urban, suburban, and rural communities from three US Midwest counties through a stochastic model that simulates the diffusion of information over time in disaster and in normal situations. To understand information diffusion in communities, we investigate the interplay of information that individuals get from online social networks, local news, government sources, mainstream media, and print media. We utilize survey data collected from target communities and create graphs of each community to quantify node-to-node and source-to-node interactions, as well as trust patterns. Monte Carlo simulation results show the average time it takes for information to propagate to 90% of the population for each community. We conclude that rural, suburban, and urban communities have different inherent properties promoting the varied flow of information. Also, information sources affect information spread differently, causing degradation of information speed if any source becomes unavailable. Finally, we provide insights on the optimal investments to improve disaster communication based on community features and contexts.
RESUMEN
The limited transparency of the inner decision-making mechanism in deep neural networks (DNN) and other machine learning (ML) models has hindered their application in several domains. In order to tackle this issue, feature attribution methods have been developed to identify the crucial features that heavily influence decisions made by these black box models. However, many feature attribution methods have inherent downsides. For example, one category of feature attribution methods suffers from the artifacts problem, which feeds out-of-distribution masked inputs directly through the classifier that was originally trained on natural data points. Another category of feature attribution method finds explanations by using jointly trained feature selectors and predictors. While avoiding the artifacts problem, this new category suffers from the Encoding Prediction in the Explanation (EPITE) problem, in which the predictor's decisions rely not on the features, but on the masks that selects those features. As a result, the credibility of attribution results is undermined by these downsides. In this research, we introduce the Double-sided Remove and Reconstruct (DoRaR) feature attribution method based on several improvement methods that addresses these issues. By conducting thorough testing on MNIST, CIFAR10 and our own synthetic dataset, we demonstrate that the DoRaR feature attribution method can effectively bypass the above issues and can aid in training a feature selector that outperforms other state-of-the-art feature attribution methods. Our code is available at https://github.com/dxq21/DoRaR.