Tensorizing neural networks
Deep neural networks currently demonstrate state-of-the-art performance in several domains.At the same time, models of this class are very demanding in terms of computational resources. In particular, a large amount of memory is required by commonly used fully-connected layers, making it hard to use the models on low-end devices and stopping the further increase of the model size. In this paper we convert the dense weight matrices of the fully-connected layers to the Tensor Train format such that the number of parameters is reduced by a huge factor and at the same time the expressive power of the layer is preserved.In particular, for the Very Deep VGG networks we report the compression factor of the dense weight matrix of a fully-connected layer up to 200000 times leading to the compression factor of the whole network up to 7 times.
Nowadays there is a large amount of demographic data which should be analysed and interpreted. From accumulated demographic data, more useful information can be extracted by applying modern methods of data mining. The aim of this study is to compare the methods of classification of demographic data by customizing the SVM kernels using various similarity measures. Since demographers are interested in sequences without discontinuity, formulas for such sequences similarity measures were derived. Then they were used as kernels in the SVM method, which is the novelty of this study. Recurrent neural network algorithms, such as SimpleRNN, GRU and LSTM, are also compared. The best classification result with SVM method is obtained using a special kernel function in SVM by transforming sequences into features, but recurrent neural network outperforms SVM.
In this paper, a deep learning method study is conducted to solve a new multiclass text classification problem, identifying user interests by text messages. We used an original dataset of almost 90 thousand forum text messages, labeled for ten interests. We experimented with different modern neural network architectures: recurrent and convolutional, as well as simpler feedforward networks. Classification accuracy was evaluated for different architectures, text representations, and sets of miscellaneous parameters.
This book constitutes the proceedings of the 7th International Conference on Analysis of Images, Social Networks and Texts, AIST 2018, held in Moscow, Russia, in July 2018.
The 29 full papers were carefully reviewed and selected from 107 submissions (of which 26 papers were rejected without being reviewed). The papers are organized in topical sections on natural language processing; analysis of images and video; general topics of data analysis; analysis of dynamic behavior through event data; optimization problems on graphs and network structures; and innovative systems.
This 2-volume set constitutes the refereed proceedings of the 9th Iberian Conference on Pattern Recognition and Image Analysis, IbPRIA 2019, held in Madrid, Spain, in July 2019.
The 99 papers in these volumes were carefully reviewed and selected from 137 submissions. They are organized in topical sections named:
Part I: best ranked papers; machine learning; pattern recognition; image processing and representation.
Part II: biometrics; handwriting and document analysis; other applications.
The main problems and features of combined approach to the complex objects control and management stability analysis are investigated in the paper. Analytical-simulation scenarios and scenarios of intelligent models and systems execution for complex objects control and management stability analysis are given. The paper describes a particular group of models and modelling systems – hybrid intelligent models and systems that allow in conditions of uncertainty, incomplete initial data and complex interdependence between elements of complex objects to evaluate the implications of realization of various scenarios and risk evaluation. The investigations have shown successful possibility of risks evaluation by the combined implementation of the analytical-simulation models and algorithms, and ANFIS method – the method of hybrid neural-fuzzy modelling.The main problems and features of combined approach to the complex objects control and management stability analysis are investigated in the paper. Analytical-simulation scenarios and scenarios of intelligent models and systems execution for complex objects control and management stability analysis are given. The paper describes a particular group of models and modelling systems – hybrid intelligent models and systems that allow in conditions of uncertainty, incomplete initial data and complex interdependence between elements of complex objects to evaluate the implications of realization of various scenarios and risk evaluation. The investigations have shown successful possibility of risks evaluation by the combined implementation of the analytical-simulation models and algorithms, and ANFIS method – the method of hybrid neural-fuzzy modelling.
One of the main objectives of strategic management is the development and selection of strategies to achieve the desired results. The main goal of this paper is the analysis of the main domains or areas of machine learning application to support the process of strategic planning and decision making. The scientific methodology of the research studies is methods and procedures of modeling and intelligent analysis. This is theoretical and empirical paper in equal measure. This paper deals with the issues of machine learning implementation and how intellectual models and systems can be used to support the process of strategic planning in the context of theory of economic growth and development. At the preprocessing stage on the basis of a modeled base of examples of strategy options, the use of clustering methods for forming groups of similar parameters that influence the choice of strategies and groups of similar enterprise objects, each of which has a certain type of strategy, are demonstrated. On the next step the selection of ranked characteristics that affect the choice of strategy is made. At the stage of solving the problem of choosing strategies, neural network and neuro-fuzzy approaches are used. The advantage of this hybrid method is based on the fact that the hybrid technology can combine the advantages of neural networks as well as the advantages of fuzzy logic.
This book constitutes the post-conference proceedings of the 4th International Conference on Machine Learning, Optimization, and Data Science, LOD 2018, held in Volterra, Italy, in September 2018.The 46 full papers presented were carefully reviewed and selected from 126 submissions. The papers cover topics in the field of machine learning, artificial intelligence, reinforcement learning, computational optimization and data science presenting a substantial array of ideas, technologies, algorithms, methods and applications.
This paper surveys some parts of approximation theory for functions of one and several real variables. Approximation of functions by algebraic polynomials is a classical theory. The paper contains some results from this theory. First results on approximation of functions by neural networks and fuzzy systems have appeared as responses on practical requirements. It was necessary to know is it possible to approximate an arbitrary continuous function by such aggregates. Later, these fields were developed like the theory of approximation of functions by algebraic polynomials. In this paper we consider some results on approximation of functions by neural networks and fuzzy systems.
In Tomsk University of Control Systems and Radioelectronics (TUSUR) one of the main areas of research is information security. The work is carried out by a scientific group under the guidance of Professor Shelupanov. One of the directions is the development of a comprehensive approach to assessing the security of the information systems. This direction includes the construction of an information security threats model and a protection system model, which allow to compile a complete list of threats and methods of protection against them. The main directions of information security tools development are dynamic methods of biometrics, methods for generating prime numbers for data encryption, steganography, methods and means of data protection in Internet of Things (IoT) systems. The article presents the main results of research in the listed areas of information security. The resultant properties in symmetric cryptography are based on the properties of the power of the generating functions. The authors have obtained symmetric principles for the development of primality testing algorithms, as discussed in the Appendix.