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.
The issues of information support based on the use of artifical neural networks for the rapid recognition of odors using devices such as "elecronic nose" are considered. The variants the reducing the test sampl for an artifical neural network are proposed with the aim of increasing the stabilutyof computatijns and the speed of calculations. A method for the rapid recognition of odors in the presence of background odors is proposed.
This book constitutes the refereed proceedings of the 6th IAPR TC3 International Workshop on Artificial Neural Networks in Pattern Recognition, ANNPR 2014, held in Montreal, QC, Canada, in October 2014. The 24 revised full papers presented were carefully reviewed and selected from 37 submissions for inclusion in this volume. They cover a large range of topics in the field of learning algorithms and architectures and discussing the latest research, results, and ideas in these areas.
We consider the technology of information processing and the use of neural networks to identify complex gas-air mixtures with a multi-sensor system of the “electronic nose” type equipped with semiconductor gas sensitive sensors. We also present the results of our experimental studies of the recognition of various gas mixtures based on the application of neural networks in the process of processing signals from a multi-sensor system of gas-sensitive sensors.
In this work, predictive scoring model for banks that estimates the data of the borrower and predicts the probabilityof a credit default by client. The goal of the work is to develop and implement a model of credit scoring using neural networks. The model will predict the solvency of bank customers. In the process of this work, the process of constructing scoring maps was studied, the problem of binary classification and some methods of solution were considered. A program has been developed in the Python programming language for assessing the solvency of bank borrowers based on the model described in the work using neural networks. In the description of the model, a detailed analysis and correction of the input data was carried out, the model was implemented taking into account the structure, training, and testing of the neural network. The result of the work has become a binary classifier, which can be used in many different tasks, where it is required to divide respondents into two classes by some criteria.
This book constitutes the proceedings of the 23rd International Symposium on Foundations of Intelligent Systems, ISMIS 2017, held in Warsaw, Poland, in June 2017. The 56 regular and 15 short papers presented in this volume were carefully reviewed and selected from 118 submissions. The papers include both theoretical and practical aspects of machine learning, data mining methods, deep learning, bioinformatics and health informatics, intelligent information systems, knowledge-based systems, mining temporal, spatial and spatio-temporal data, text and Web mining. In addition, four special sessions were organized; namely, Special Session on Big Data Analytics and Stream Data Mining, Special Session on Granular and Soft Clustering for Data Science, Special Session on Knowledge Discovery with Formal Concept Analysis and Related Formalisms, and Special Session devoted to ISMIS 2017 Data Mining Competition on Trading Based on Recommendations, which was launched as a part of the conference.
The medium-term forecasting of the sea ice extent has been carried out by determining of the relationship between incoming solar radiation and the sea ice extent in the Northern Hemisphere. Different methods of the statistic and neural modeling have been used. Forecast shows that the main factor determining the variation of the maximum and minimum sea ice extent in the medium-term scale is the variability of solar radiation arrived at the top of the atmosphere. Evaluation of the medium-term forecasts of the sea ice extent demonstrates the effectiveness of using the averaged results of the regression analysis and neural network modeling.
The first part of the issue gives general information about foreign exchange market (FOREX), review of forecasting foreign exchange rate is given. In addition we will consider the new model of nonlinear analysis to give a broader theoretical basis to the research - an artificial neural network (ANN).The nonlinear analysis and the ANN is still improving, but even now the researchers notice evident advantages of the method.
Сurrent state of risk management allows innovative companies to use a variety of tools to identify, analyze and assess various risk factors. The article describes the application of neural net technology for the elimination of key discovered risks.
The secular outcome of our investigation is development of new monitoring service for glucose control related to diabetes. It is based on the main results of research: 1) New innovative wearable sensor that carry non-invasive measurement of glucose level. Sensor uses several independent technologies, simultaneously: radio-frequency with different levels of signal, ultrasonic, electromagnetic and thermal; 2) Special mobile application as the principal interface monitor for personal usage; 3) The unique proprietary algorithm, based on neural net, which calculates the weighted average and returns the user's glucose level. The algorithm précises the results of measurements, based on genetic neural net ideas; 4) Special designed Data Base Storage in our cloud software specialized for gathering information and giving the predictions for patient. All results together makes the essence of the research.