This paper presents a comprehensive approach to SAR image retrieval for navigation in GPS-denied areas. It explores the utilization of SAR images to develop navigation techniques, assuming the system can generate real-time SAR images. The navigation process involves image retrieval, where a query image is compared to stored images in a database to identify the most similar ones. The selected images serve as reference points for extracting precise location coordinates. We propose model leveraging on the notion of siamese artificial neural networks, inspired by the SqueezeNet architecture, that incorporates swish activation functions and a retrieval module to address challenges related to variations in UAV height and rotations. Extensive evaluation demonstrates the effectiveness of the proposed method, with low retrieval error and feasibility for computationally constrained devices. The stability of the method is also validated using out-of-sample data. Overall, this work contributes to the advancement of SAR image retrieval and navigation in GPS-denied environments, with potential applications in navigation, target detection, terrain classification, and more.

The monitoring of data streams with a network structure have drawn increasing attention due to its wide applications in modern process control. In these applications, high-dimensional sensor nodes are interconnected with an underlying network topology. In such a case, abnormalities occurring to any node may propagate dynamically across the network and cause changes of other nodes over time. Furthermore, high dimensionality of such data significantly increased the cost of resources for data transmission and computation, such that only partial observations can be transmitted or processed in practice. Overall, how to quickly detect abnormalities in such large networks with resource constraints remains a challenge, especially due to the sampling uncertainty under the dynamic anomaly occurrences and network-based patterns. In this paper, we incorporate network structure information into the monitoring and adaptive sampling methodologies for quick anomaly detection in large networks where only partial observations are available. We develop a general monitoring and adaptive sampling method and further extend it to the case with memory constraints, both of which exploit network distance and centrality information for better process monitoring and identification of abnormalities. Theoretical investigations of the proposed methods demonstrate their sampling efficiency on balancing between exploration and exploitation, as well as the detection performance guarantee. Numerical simulations and a case study on power network have demonstrated the superiority of the proposed methods in detecting various types of shifts. Note to Practitioners—Continuous monitoring of networks for anomalous events is critical for a large number of applications involving power networks, computer networks, epidemiological surveillance, social networks, etc. This paper aims at addressing the challenges in monitoring large networks in cases where monitoring resources are limited such that only a subset of nodes in the network is observable. Specifically, we integrate network structure information of nodes for constructing sequential detection methods via effective data augmentation, and for designing adaptive sampling algorithms to observe suspicious nodes that are likely to be abnormal. Then, the method is further generalized to the case that the memory of the computation is also constrained due to the network size. The developed method is greatly beneficial and effective for various anomaly patterns, especially when the initial anomaly randomly occurs to nodes in the network. The proposed methods are demonstrated to be capable of quickly detecting changes in the network and dynamically changes the sampling priority based on online observations in various cases, as shown in the theoretical investigation, simulations and case studies.

The elasticity principles have become one of the most important features of contemporary distributed systems. This novel paradigm has major consequences on Petri nets modeling. The problem is to decide whether an arbitrary Petri net satisfies the elasticity and how to make it elastic. The problem is solved by synthesis of Fractal Petri nets. Fractal Petri nets based on Iterated Algebraic System which makes possible elastic synthesis by incrementing or decrementing of original Petri net. One of the important tasks is recognition of a subnet which used as a pattern for elastic synthesis. For this purpose evolutionary classification of Petri nets based on subnets synthesis is introduced. And furthermore, algebraic operations replication and composition make elastic synthesis object-oriented. Control of the elasticity over the resources and system behavior is achieved with the use of patterns overlapping technique.

Most recommender algorithms are designed to suggest relevant items, but suggesting these items does not always result in user satisfaction. Therefore, the efforts in recommender systems recently shifted towards serendipity, but generating serendipitous recommendations is difficult due to the lack of training data. To the best of our knowledge, there are many large datasets containing relevance scores (relevance oriented) and only one publicly available dataset containing a relatively small number of serendipity scores (serendipity oriented). This limits the learning capabilities of serendipity oriented algorithms. Therefore, in the absence of any known deep learning algorithms for recommending serendipitous items and the lack of large serendipity oriented datasets, we introduce SerRec our novel transfer learning method to recommend serendipitous items. SerRec uses transfer learning to firstly train a deep neural network for relevance scores using a large dataset and then tunes it for serendipity scores using a smaller dataset. Our method shows benefits of transfer learning for recommending serendipitous items as well as performance gains over the state-of-the-art serendipity oriented algorithms

This paper presents a Pattern-type Reachability Analysis of distributed systems modelled by Fractal Petri Nets. We define Fractal Petri Nets as a net which is dynamically synthesized from self-similar Place/Transition subnets on the base of algebraic operations. Fractal Petri nets are considered as distributed. This means, that every self-similar subnet can be considered autonomously. Resources can be shared between subnets of Fractal net. For purpose of modelling shared resources the operation of overlapping places of the subnets is introduced. To investigate dynamic properties of Fractal Petri Nets a Pattern-type Reachability Analysis is developed. It is helpful for understandability and readability of complicated reachability trees.

During the last decade a new kind of dynamic architectural models and algorithms for system scalability has emerged. The dynamic scalability is one of the most important design goals for developers of distributed systems and cloud computing. We introduce Fractal Petri nets (FP-nets) that takes into account the dynamically scalable distributed architectures. FP-nets are a backward compatible extension of Petri Nets on the base of algebraic structure. FP-nets are synthesized from the initial marking Petri net by replication. FP-nets are considered as distributed. Well-known techniques for analysis of Petri net theory are adjusted to FP-nets. New techniques for modeling distributed resources named “token-stub” are suggested. The distributed IT resources are integrated into the architectures of FP-nets. Three FP-nets architectures such as layered, folded, and unfolded are proposed. A FP-net avoids the scalability problems. There are properties of FP-nets: scalability, self-similarity, and token-stub dependences. FP-nets can be highly recommended for automatic modeling dynamically scalable distributed architectures.

Nowadays there are many social media sites with a very large number of users. Users of social media sites and relationships between them can be modelled as a graph. Such graphs can be analysed using methods from social network analysis (SNA). Many measures used in SNA rely on computation of shortest paths between nodes of a graph. There are many shortest path algorithms, but the majority of them suits only for small graphs, or work only with road network graphs that are fundamentally different from social graphs. This paper describes an efficient shortest path searching algorithm suitable for large social graphs. The described algorithm extends the Atlas algorithm. The proposed algorithm solves the shortest path problem in social graphs modelling sites with over 100 million users with acceptable response time (50 ms per query), memory usage (less than 15 GB of the primary memory) and applicable accuracy (higher than 90% of the queries return exact result).

Social media sites have appeared during the last 10 years and their use has exploded all over the world. Twitter is a microblogging service that has currently 320 million user profiles and over 100 million daily active users. Many celebrities and leading politicians have a verified profile on Twitter, including Justin Bieber, president Obama, and the Pope. In this paper we investigate the '‘hundreds of Putins and Obamas phenomenon’ on Twitter. We collected two data sets in 2015 containing 582 and 6477 profiles that are related to the G20 leaders’ profiles on Twitter. The number of namesakes varied from 5 to 1000 per leader. We analysed in detail various aspects of the Putin and Erdogan related profiles. For the first ones we looked into the language of the profiles, their follower sets, the address in the profile and where the tweets were really sent from. For both profile sets we investigated why the accounts were created. For this, we deduced 12 categories based on the information in the profile and the contents of the sent tweets. The research is exploratory in nature, but we tentatively looked into online identity, communication and political theories that might explain emergence of these kinds of Twitter profiles.

In this article we describe a framework (DEGA-Gen) for the application of distributed genetic algorithms for detection of communities in networks. The framework proposes efficient ways of encoding the network in the chromosomes, greatly optimizing the memory use and computations, resulting in a scalable framework. Different objective functions may be used for producing division of network into communities. The framework is implemented using open source implementation of MapReduce paradigm, Hadoop. We validate the framework by developing community detection algorithm, which uses modularity as measure of the division. Result of the algorithm is the network, partitioned into non-overlapping communities, in such a way, that network modularity is maximized. We apply the algorithm to well-known data sets, such as Zachary Karate club, bottlenose Dolphins network, College football dataset, and US political books dataset. Framework shows comparable results in achieved modularity; however, much less space is used for network representation in memory. Further, the framework is scalable and can deal with large graphs as it was tested on a larger youtube.com dataset.

This paper describes traces of user activity around a alleged online social network profile of a Boston Marathon bombing suspect, after the tragedy occurred. The analyzed data, collected with the help of an automatic social media monitoring software, includes the perpetrator's page saved at the time the bombing suspects' names were made public, and the subsequently appearing comments left on that page by other users. The analyses suggest that a timely protection of online media records of a criminal could help prevent a large-scale public spread of communication exchange pertaining to the suspects/criminals' ideas, messages, and connections.

Nowadays, WWW contains a number of social media sites, which are growing rapidly. One of the main features of social media sites is to allow to its users creation and modification of contents of the site utilizing the offered WWW interfaces. Such contents are referred to as user generated contents and their type varies from site to site. Social media sites can be modeled as constantly evolving multirelational directed graphs. In this paper we discuss persistent data structures for such graphs, and present and analyze queries performed against the structures. We also estimate the space requirements of the proposed data structures, and compare them with the naive "store each complete snapshot of the graph separately". We also investigate query performance against our data structure. We present analytical estimation results, simulation results, and discuss its performance when it is used to store entire contents of Live journal.