Several recent works have shown that image descriptors produced by deep convolutional neural networks provide state-of-the-art performance for image classification and retrieval problems. It has also been shown that the activations from the convolutional layers can be interpreted as local features describing particular image regions. These local features can be aggregated using aggregation approaches developed for local features (e.g. Fisher vectors), thus providing new powerful global descriptors. In this paper we investigate possible ways to aggregate local deep features to produce compact global descriptors for image retrieval. First, we show that deep features and traditional hand-engineered features have quite different distributions of pairwise similarities, hence existing aggregation methods have to be carefully re-evaluated. Such re-evaluation reveals that in contrast to shallow features, the simple aggregation method based on sum pooling provides arguably the best performance for deep convolutional features. This method is efficient, has few parameters, and bears little risk of overfitting when e.g. learning the PCA matrix. Overall, the new compact global descriptor improves the state-of-the-art on four common benchmarks considerably.
The article is dedicated to the method of aggregation of financial analysts’ recommendations in the framework of the evidence theory. This method considered on the example of Russian stock market and the quality of the obtained results was compared with the classical consensus forecast. It is shown that the combination rules, which are widely developed in the theory of evidence, allow aggregating the recommendations of analysts taking into account the historical reliability of information sources, the nature of the taken decisions (pessimism-optimism), the conflict between forecasts and recommendations, etc. In most cases it turned out that, obtained aggregated forecasts are more accurate than consensus forecast.
Word-formation processes in Aghul (a Northeast Caucasian language spoken in Daghestan, Russia) include both compounding and derivation. Verbal compounding is very productive and is the primary way of enriching the verbal lexicon in the modern language, using borrowed Russian verbs. In contrast, although there are quite a large number of nominal compounds, they seem to be fixed expressions and no new compounds are created. Derivation is mainly suffixal with the exception of verbal locative and repetitive derivation achieved by prefixes. Various types of full reduplication, as well as echo-reduplication and partial reduplication are fairly productive.
Modern companies continue investing more and more in the creation, maintenance and change of software systems, but the proper specification and design of such systems continues to be a challenge. The majority of current approaches either ignore real user and system runtime behavior or consider it only informally. This leads to a rather prescriptive top-down approach to software development. In this paper, we propose a bottom-up approach, which takes event logs (e.g., trace data) of a software system for the analysis of the user and system runtime behavior and for improving the software. We use well-established methods from the area of process mining for this analysis. Moreover, we suggest embedding process mining into the agile development lifecycle. The goal of this position paper is to motivate the need for foundational research in the area of software process mining (applying process mining to software analysis) by showing the relevance and listing open challenges. Our proposal is based on our experiences with analyzing a big productive touristic system. This system was developed using agile methods and process mining could be effectively integrated into the development lifecycle.
Apart from the public sphere and the norms set by society, the private sphere plays an important role in the lives of the disabled, including the personal experience of disability at a micro level: in their families, everyday routines and romantic relationships. In this chapter, issues of family structure are considered using a narrative analysis of interviews with women who use wheelchairs. Various cultural, social, economic and political determinants effect the formation of certain types of family structure and attitudes towards family life. At the same time, they interrelate with biographical factors that reinforce or weaken the limits of freedom and private life. Using narrative analysis, I demonstrate what role family plays in constructing the identity of a person with a disability, and how family members act as coauthors of individual biographies. This can be seen in those dilemmas of family life associated with the feelings, sexuality and emotional stability at the micro-level of the life experience and identification of women with disabilities.
Статья посвящена описанию феномена агитпропа в современном искусстве.
The philosophical and scientific debate on the limits of knowledge arose and took place mainly in the German and Anglo-Saxon cultural area; this fact notwithstanding, it has to be acknowledged that Italy, as well, significantly contributed to its spread and critical development. This paper is focussed on the most relevant Italian milieus, figures and themes relating to the debate on the limits of knowledge.
In this paper, we present a modification of dynamic programming algorithms (DPA), which we denote as graphical algorithms (GrA). For some single machine scheduling problems, it is shown that the time complexity of the GrA is less than the time complexity of the standard DPA. Moreover, the average running time of the GrA is often essentially smaller. A GrA can also solve large-scale instances and instances, where the parameters are not integer. For some problems, GrA has a polynomial time complexity in contrast to a pseudo-polynomial complexity of a DPA.
ASM, informal mining
The article is devoted to the analysis of the semantics of the word stereoma. The Septuagint as it was understood by a Greek rhetorician: Pseudo-Longinus and στερέωμα. The paper deals with the first (and only) quotation from the Bible in the classical Greek literature: a quotation from the opening chapter of Genesis in a treatise on eloquence, Περὶ Ὕψους, written presumably in the first century CE by an anonymous Greek author, commonly referred to as Pseudo-Longinus. One can see at a gl ance that the wording of the quotation differs considerably from that of the Greek Genesis. We suggest that the difference is due to the wrong understanding of Gen 1:6 by the author of Περὶ Ὕψους. The present paper attempts to reconstruct how a Greek rhetorician, experienced in classical literature but not versed in the Bible, could understand and interpret the biblical account of the creation of the Heavens, especially the word στερέωμα “solid body” used in the Greek Bible (Gen 1:6) in the meaning “heaven”. This meaning is a neologism coined by the authors of the Septuagint. The paper shows, with a reference to the classical literature and Basil the Great (Hexaemeron), that the word στερέωμα would seem to a Greek rhetorician as a much more appropriate designation for the Earth than for the Heaven. It also shows that what was said about the στερέωμα in Gen 1:6 would also point in the same direction. The biblical Στερέωμα ἐν μέσῳ τοῦ ὕδατος – “a solid body in the midst of the waters” – could not have been understood by a Greek philosopher or rhetorician as “the Heaven”. One may rather suppose, it must have been understood as “the Earth”. If we assume that Pseudo-Longinus borrowed the quotation of Gen 1:6 from some source without knowing its wider context, we shall be able to explain how the wording of Περὶ Ὕψους emerged from that of the Septuagint: as a result of misreading caused by linguistic and cultural differences between the world of the Greek-speaking Jews and that of classical antiquity.
We propose deterministic and stochastic models of clock synchronization in nodes of large distributed network locally coupled with a reliable external exact time server.
Stochastic Local Search (SLS) is one of the most popular approaches to Boolean satisfiability problem and solvers based on this algorithm have made a substantial progress over the years. However, nearly all state of the art SLS solvers do not attempt to find a good starting point, instead using random values. We present a heuristic for finding an initial assignment based on non-linear optimization of continuous extension of given Boolean formula. This heuristic works by optimizing continuous function that represents the formula and then converting the result into discrete values. We also provide experimental evaluation of new heuristic implemented in ProbSAT solver.
The Vehicle Routing Problem (VRP) is one of the most popular combinatorial optimization problems which is closely related to the real-life optimization challenges. Being developed for more than 60 years the problem has been considered in many different formulations. In real-life goods distribution such constraints as fleet size and mix, sitedependency constraints, hard and soft time windows, vehicle capacity constraints are very important. In this paper we consider Capacitated Vehicle Routing Problem with hard Time Windows. We propose a hybrid heuristic algorithm which contains elements of ant colony optimization strategy and tabu search technique. Our algorithm shows good performance and results for the well-known Solomon dataset.
In this paper we observe the opportunity to offer new methods of solving NP-hard problems which frequently arise in the domain of information management, including design of database structures and big data processing. In our research we are focusing on the Maximum Clique Problem (MCP) and propose a new approach to solving that problem. The approach combines the artificial neuro-network paradigm and genetic programming. For boosting the convergence of the Hopfield Neural Network (HNN) we propose the genetic algorithm as the selection mechanism for terms of energy function. As a result, we demonstrate the proposed approach on experimental graphs and formulate two hypotheses for further research.
The paper discusses a multi-paradigm approach to the modeling of Demand Responsive Transport systems. It contains a brief overview of issues which appear during modeling of such systems, considers various multi-agent architectures and describes some algorithms which can be used for modeling. Also the paper provides some details about previous investigations on this topic, in particular: a centralized model based on combinatorial auctions and a multi-agent based multi-layer distributed hybrid model. The aim of the paper is working out a sound solution based on a combination of these two approaches which would utilize “system of systems” engineering approach where layered architecture would help to deal with real-time issues and increase system’s reliability and combinatorial auctions would help with global search of the optimal solution. Such combination improves the efficiency and reliability of the system.
In this paper, we provide the solution for RecSys Challenge 2018 by our Avito team, which obtained the 3rd place in main track. The goal of the competition was to recommend music tracks for automatic playlist continuation. As a part of this challenge, Spotify released a large public dataset, which allowed us to train a rather complex algorithm. Our approach consists of two stages: collaborative filtering for candidate selection and gradient boosting for final prediction. The combination of these two models performed well with the playlist and track metadata given.
In 1937, the Japanese economist Kaname Akamatsu discovered specific links between the rise and decline of the global peripheries. Akamatsu’s theory of development describes certain mechanisms whose working results in the narrowing of the gap between the level of development of the economy of developing and developed countries, and, thus, in the re-structuring of the relationships between the global core and the global periphery. Akamatsu developed his model on the basis of his analysis of the economic development of Japan before World War II, with a special emphasis on the development of the Japanese textile industry. Akamatsu’s catch-up development includes three phases: import of goods, organization of the production of previously imported products, and export of those goods. This model proved to be productive for analyzing the development of many other developing countries, especially in East Asia, making the theory of flying geese popular among the economists of these countries, as well as the whole world. The “flying geese” model produces certain swings that may be denoted as Akamatsu waves. Akamatsu waves may be defined as cycles (with a period ranging from 20 to 60 years) that are connected with convergence and divergence of core and periphery of the World System in a way that explains cyclical upward and downward swings (at global and national levels) in the movements of the periphery countries as they catch up with the richer ones.