In recent years, taking into account the development of the needs of human society, a great need has arisen for the automation of various processes. To solve such problems, are used neural network technologies, which are attempts to reproduce the human nervous system, are actively gaining great popularity. In particular, neural networks are actively used in the tasks of classifying objects on images, and therefore, based on this, the task of developing new or modifying existing algorithms is relevant. This article presents an analysis of existing algorithms for using neural systems to classify objects in images, as well as technologies for constructing machine vision systems. An attempt is made to develop the architecture of a system for detecting objects on images, is capable of scaling both vertically and horizontally, which is characterized by sufficient accuracy and speed. Based on the developed architecture, a system for automating the process of classifying objects in images was developed. To give recommendations, a software module was developed that uses machine learning to identify the rules by which the class of the image object should be made. Neural models implementing the classification of objects in images were also configured and trained. The classification results are checked on a real data set, which is similar in composition of attributes to that which will be during production use of the system.
The high-fidelity image generation has been a subject of active research in the recent past. It provides benchmark towards image decoder’s performance. The autoregressive image models generate small images successfully but scalability has been a problem. The challenges include vast encoding of previous context and learning distribution which maintains global semantic coherence and exactness. These issues have been addressed through subscale pixel network (SPN) and multidimensional upscaling. To improve the accuracy further, in this work a hierarchical version of image generation model is presented. It disentangles background, object shape and appearance to hierarchically generate images of fine-grained object categories. To achieve this information theory associates a factor to latent code and condition relationships between codes to induce hierarchy. The hierarchical model’s learned features are used to cluster real images. The experimental results on ImageNet and CelebAHQ datasets for different image sizes highlight hierarchical model’s superiority against the benchmarks. The images are generated with better fidelity with respect to large scale samples.
Early warning systems are essential to mitigate the consequences of outbreaks of infectious diseases, which causes millions of deaths every year. Surveillance systems collect data on epidemic-prone diseases to trigger prompt public health interventions. In recent years, these systems have improved significantly thanks to infodemiology, a recent research field that promotes the use of health-data collected from the Internet. However, early warning systems can be improved regarding the interpretability and confidentiality of the compiled evidences. In addition, they still rely heavily on human intervention to distinguish among confident and non-confident evidences. To solve these concerns we present CollaborativeHealth, an infodemiology platform that compiles evidence from (1) social networks, (2) public reports, and (3) voluntary citizen participation; and makes use of deep-learning technologies to extract knowledge regarding infectious diseases, their symptoms, or poor environment conditions what promote the propagation of these diseases. Finally, all the compiled evidence is available to health-professionals in real-time through a configurable dashboard. The validation of CollaborativeHealth was performed with a real use-case about monitoring infectious disease cases related to Zika, Dengue, Chikungunya, and Influenza in Ecuador.
The aim of this paper is to define a system built on a modular principle and to construct a recursive algorithm for calculating its reliability. For this purpose, the system is described using a monotone Boolean function that characterizes its performance. The oriented tree is mapped to a Boolean function (which characterizes operability). From any leaf of this tree there is a (single) path to its root. We assume that all the edges of this path are directed towards the root. A recursive algorithm for calculating the Boolean function and the reliability of the system, described by this function, is constructed. It is proved that the number of arithmetic operations for calculating a Boolean function, built on the modular principle, and the reliability of the system depend linearly on the number of leaves in the tree. Then how for systems of general form of this dependence is exponential. #CSOC1120.
Information security data breaches are becoming larger and more frequent. Incorporating information security into the culture of the information technology (IT) staff members that support these technologies is a key function that must be considered in parallel to improved security technology. The framework proposed in this paper considers focusing on cost-reducing products, services and structures while building the correct behaviour and values in IT staff members and strengthening their ability to improve information security assessment capabilities in the organization to better support information security management. A tool to evaluate the framework is also described as well as concise feedback on how the framework and tool was tested in a few organizations.
