The Internet of Things paradigm is creating an environment where the big data originators will be located at the edge of the Internet. Accordingly, data analytic infrastructure is also being relocated to the network edges, to fulfill the philosophy of data gravity, under the umbrella of Fog Computing. The extreme edge of the hierarchical infrastructure consists of sensor devices that constitute the Wireless Sensor Networks. The role of these devices has evolved tremendously over the past few years owing to significant improvements in their design and computational capabilities. Sensor devices, today, are not only capable of performing sense and send tasks but also certain kinds of in-network processing. As such, triple optimization of sensing, computing and communication tasks is required to facilitate the implementation of data analytics on the sensor devices. A sensor node may optimally partition a computation task, for instance, and offload sub-tasks to cooperative neighbouring nodes for parallel execution to, in turn, optimize the network resources. This approach is crucial, especially, for energy harvesting sensor devices where the energy profile and, therefore, the computation capability of each device differs depending on the node location and time of day. Accordingly, future in-network computing must capture the energy harvesting information of sensor nodes to jointly optimize the computation and communication within the network. In this paper, we present a theoretical model for computation offloading in micro-solar powered energy harvesting sensor devices. Optimum data partitioning to minimize the total energy consumption has been discussed based on the energy harvesting status of sensor nodes for different scenarios. The simulation results show that our model reduced both energy losses and waste due to energy conversion and overflows respectively compared to a data partitioning algorithm that offloads computation tasks without taking the energy harvesting status of nodes into consideration. Our approach also improves energy balance of a WSN which is an important factor for its long-term autonomous operation.
Distributed analytics facilitate to make the data-driven services smarter for a wider range of applications in many domains, including agriculture. The key to producing services at such level is timely analysis for deriving insights from reliable data. Centralized data analytic services are becoming infeasible due to limitations in the Information and Communication Technologies (ICT) infrastructure, timeliness of the information, and data ownership. Distributed Machine Learning (DML) platforms facilitate efficient data analysis and overcome such limitations effectively. Federated Learning (FL) is a DML methodology, which enables optimizing resource consumption while performing privacy-preserved timely analytics. In order to create such services through FL, there needs to be innovative machine learning (ML) models as data complexity as well as application requirements limit the applicability of existing ML models. Even though NN-based models are highly advantageous, use of NN in FL settings is limited with thin clients (with less computational capabilities) and high-dimensional data (with large number of model parameters). Therefore, in this paper, we propose a novel Neural Network (NN)and Partial Least Square (PLS) regression-based joint FL model (FL-NNPLS). Its predictive performance is evaluated under sequentialand parallel-updating based FL algorithms in a smart farming context for milk quality analysis. Smart farming is a fast-growing industrial sector which requires effective analytics platforms to enable sustainable farming practices. However, the use of advanced ML techniques are still at a early stage for improving the effectiveness of farming practices. Our FL-NNPLS approach performs and compares well with a centralized approach and demonstrates state-of-the-art performance.
Respiratory viruses including Respiratory syncytial virus (RSV), influenza virus and cornaviruses such as Middle Eastern respiratory virus (MERS) and SARS-CoV-2 infect and cause serious and sometimes fatal disease in thousands of people annually. It is critical to understand virus propagation dynamics within the respiratory system because new insights will increase our understanding of virus pathogenesis and enable infection patterns to be more predictable in vivo, which will enhance targeting of vaccines and drug delivery. This study presents a computational model of virus propagation within the respiratory tract network. The model includes the generation network branch structure of the respiratory tract, biophysical and infectivity properties of the virus, as well as air flow models that aid the circulation of the virus particles. The model can also consider the impact of the immune response aim to inhibit virus replication and spread. The model was applied to the SARS-CoV-2 virus by integrating data on its life-cycle, as well as density of Angiotensin Converting Enzyme (ACE2) expressing cells along the respiratory tract network. Using physiological data associated with the respiratory rate and virus load that is inhaled, the model can improve our understanding of the concentration and spatiotemporal dynamics of virus
Social network analysis is a technique to study behavioral dynamics within a social group. In social network analysis, it is an open question as to whether it is possible to characterize animal-level behaviors by using group-level information. Also, it was believed that the combined use of social network analysis would provide a more comprehensive understanding about social dynamics. In light of these two factors, here we explain an approach to evaluate animal importance to a group by considering the variability in group-level structural information, which is computed by joining the animal-and group-level social network analysis measures node centrality and network entropy, respectively. Moreover, two other metrics, animal social interaction range and nearest-neighbor frequency matrix, which represent social affiliation of each animal within the group, are computed to help addresses the general challenges in graph-based social network analysis, and thereby, improve the precision of animal importance measures. Finally, we derive the joint distribution of animal importance of the group in detecting atypical social behaviors. The approach is tested using tracking data of dairy cows. The reliability of the derived animal importance was superior to the already existing animal importance measures. To illustrate the usability of the animal importance metric, a simulation study was conducted to identify sick and estrus animals in a group. The social affiliation of sick cows was less when compared to healthy cows. Also, their individual distributions of animal importance were shifted towards the left of the mean of the animal importance distributions of healthy cows. Consequently, the joint distribution of animal importance of the group exhibited a bi-model distribution with a left tailored shape. The behavior of cows in estrus was opposite to that of sick cows. Moreover, with increasing number of sick and estrus cows in the group, respectively the group entropy decreased with larger variance and slightly increased with less variance. Therefore, the entropy-based animal importance metric has superior performances when evaluating animal importance to the group compared to the existing metrics. It can be used
Smart dairy farming has become one of the most exciting and challenging area in cloud-based data analytics. Transfer of raw data from all farms to a central cloud is currently not feasible as applications are generating more data while internet connectivity is lacking in rural farms. As a solution, Fog computing has become a key factor to process data near the farm and derive farm insights by exchanging data between on-farm applications and transferring some data to the cloud. In this context, learning in the compressed data domain, where decompression is not necessary, is highly desirable as it minimizes the energy used for communication/computation, reduces required memory/storage, and improves application latency. Mid-infrared spectroscopy (MIRS) is used globally to predict several milk quality parameters as well as deriving many animal-level phenotypes. Therefore, compressed learning on MIRS data is beneficial both in terms of data processing in the Fog, as well as storing large data sets in the cloud. In this paper, we used principal component analysis and wavelet transform as two techniques for compressed learning to convert MIRS data into a compressed data domain. The study derives near lossless compression parameters for both techniques to transform MIRS data without impacting the prediction accuracy for a selection of milk quality traits.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.