More Like this

1. How can sharing knowledge better explain land use changes in Ethiopia?

   Cara Steger, Girma Nigussie (February 2022, Environmental science journal for teens)

   Allison Gamzon and Alexandra Appleton (Ed.)

   When you think about environmental issues, you probably think about science. How does science explain the problem? How can we use science to create a solution? Yet if we only think like scientists, we may miss important details. That is why we need to share knowledge with people who have different perspectives. We shared knowledge with local residents near the Guassa grassland in Ethiopia to discuss how land use had changed over time. They described how they use the land and the benefits it provides. Using satellite technology, we created maps of the area. We then shared these maps with the people of Guassa. Together, we found a link between soil problems and land use changes. We also determined how different management strategies affect land use. Because we shared knowledge with the local residents, we created a more complete picture of what is happening in Guassa. 

   more » « less
2. A delayed reaction–diffusion viral infection model with nonlinear incidences and cell-to-cell transmission

   https://doi.org/10.1142/S179352452150100X  

   Ge, Qing; Wang, Xia; Rong, Libin (November 2021, International Journal of Biomathematics)

   In this paper, we propose a reaction–diffusion viral infection model with nonlinear incidences, cell-to-cell transmission, and a time delay. We impose the homogeneous Neumann boundary condition. For the case where the domain is bounded, we first study the well-posedness. Then we analyze the local stability of homogeneous steady states. We establish a threshold dynamics which is completely characterized by the basic reproduction number. For the case where the domain is the whole Euclidean space, we consider the existence of traveling wave solutions by using the cross-iteration method and Schauder’s fixed point theorem. Finally, we study how the speed of spread in space affects the spread of cells and viruses. We obtain the existence of the wave speed, which is dependent on the diffusion coefficient. 

   more » « less
3. Neural Networks with Sparse Activation Induced by Large Bias: Tighter Analysis with Bias-Generalized NTK

   Yang, Hongru; Jiang, Ziyu; Zhang, Ruizhe; Liang, Yingbin; Wang, Zhangyang (November 2024, Journal of machine learning research)

   We study training one-hidden-layer ReLU networks in the neural tangent kernel (NTK) regime, where the networks' biases are initialized to some constant rather than zero. We prove that under such initialization, the neural network will have sparse activation throughout the entire training process, which enables fast training procedures via some sophisticated computational methods. With such initialization, we show that the neural networks possess a different limiting kernel which we call bias-generalized NTK, and we study various properties of the neural networks with this new kernel. We first characterize the gradient descent dynamics. In particular, we show that the network in this case can achieve as fast convergence as the dense network, as opposed to the previous work suggesting that the sparse networks converge slower. In addition, our result improves the previous required width to ensure convergence. Secondly, we study the networks' generalization: we show a width-sparsity dependence, which yields a sparsity-dependent Rademacher complexity and generalization bound. To our knowledge, this is the first sparsity-dependent generalization result via Rademacher complexity. Lastly, we study the smallest eigenvalue of this new kernel. We identify a data-dependent region where we can derive a much sharper lower bound on the NTK's smallest eigenvalue than the worst-case bound previously known. This can lead to improvement in the generalization bound. 

   more » « less
4. A Complete Study of P.K.I. (PKI’s Known Incidents)

   https://doi.org/10.2139/ssrn.3425554  

   Serrano, Nicolas; Hadan, Hilda; Camp, L. Jean (September 2019, SSRN Electronic Journal)

   null (Ed.)

   In this work, we report on a comprehensive analysis of PKI resulting from Certificate Authorities’ (CAs) behavior using over 1300 instances. We found several cases where CAs designed business models that favored the issuance of digital certificates over the guidelines of the CA Forum, root management programs, and other PKI requirements. Examining PKI from the perspective of business practices, we identify a taxonomy of failures and identify systemic vulnerabilities in the governance and practices in PKI. Notorious cases include the “backdating” of digital certificates, the issuance of these for MITM attempts, the lack of verification of a requester’s identity, and the unscrupulous issuance of rogue certificates. We performed a detailed study of 379 of these 1300 incidents. Using this sample, we developed a taxonomy of the different types of incidents and their causes. For each incident, we determined if the incident was disclosed by the problematic CA. We also noted the Root CA and the year of the incident. We identify the failures in terms of business practices, geography, and outcomes from CAs. We analyzed the role of Root Program Owners (RPOs) and differentiated their policies. We identified serial and chronic offenders in the PKI trusted root programs. Some of these were distrusted by RPOs, while others remain being trusted despite failures. We also identified cases where the concentration of power of RPOs was arguably a contributing factor in the incident. We identify these cases where there is a risk of concentration of power and the resulting conflict of interests. Our research is the first comprehensive academic study addressing all verified reported incidents. We approach this not from a machine learning or statistical perspective but, rather, we identify each reported public incident with a focus on identifying patterns of individual lapses. Here we also have a specific focus on the role of CAs and RPOs. Building on this study, we identify the issues in incentive structures that are contributors to the problems. 

   more » « less
5. Robust Estimation of Tree Structured Gaussian Graphical Models

   Katiyar, Ashish; Hoffmann, Jessica; Caramanis, Constantine (June 2019, Proceedings of Machine Learning Research)

   Consider jointly Gaussian random variables whose conditional independence structure is specified by a graphical model. If we observe realizations of the variables, we can compute the covariance matrix, and it is well known that the support of the inverse covariance matrix corresponds to the edges of the graphical model. Instead, suppose we only have noisy observations. If the noise at each node is independent, we can compute the sum of the covariance matrix and an unknown diagonal. The inverse of this sum is (in general) dense. We ask: can the original independence structure be recovered? We address this question for tree structured graphical models. We prove that this problem is unidentifiable, but show that this unidentifiability is limited to a small class of candidate trees. We further present additional constraints under which the problem is identifiable. Finally, we provide an O(n^3) algorithm to find this equivalence class of trees. 

   more » « less