Dmitry Berg, Olga Zvereva (Ural Federal University)
Identification Of Autopoietic Communication Patterns In Social And Economic Networks
AIST Conference 2015 http://aistconf.org
The document discusses building software to test other software using an excellent team equipped with good tools. It mentions working on improving the quality and efficiency of exchange platforms and expresses pride in supporting the AIST conference, as well as interest in machine learning and data mining research.
Artem Lukanin - Normalization of Non-Standard Words with Finite State Transd...AIST
This document discusses text normalization for Russian speech synthesis. It introduces Normatex, an open-source Russian text normalization system using finite state transducers. Normatex expands non-standard words like numbers, abbreviations, and acronyms. It achieved 84.33% recall and 93.95% precision on a test corpus. The document provides details on Normatex's normalization of numbers, acronyms, abbreviations, and its finite state transducers. Further improvements to Normatex are still underway.
The document discusses building software to test other software using an excellent team equipped with good tools. It mentions working on improving the quality and efficiency of exchange platforms and expresses pride in supporting the AIST conference, as well as interest in machine learning and data mining research.
Artem Lukanin - Normalization of Non-Standard Words with Finite State Transd...AIST
This document discusses text normalization for Russian speech synthesis. It introduces Normatex, an open-source Russian text normalization system using finite state transducers. Normatex expands non-standard words like numbers, abbreviations, and acronyms. It achieved 84.33% recall and 93.95% precision on a test corpus. The document provides details on Normatex's normalization of numbers, acronyms, abbreviations, and its finite state transducers. Further improvements to Normatex are still underway.
Nataly Zhukova - Conceptual Model for Routine Measurements Analyses in Seman...AIST
The document presents a conceptual model for processing routine measurements in adaptive intelligent information systems. The model aims to help users solve domain-specific tasks by reducing large amounts of data, linking related pieces of information, and providing machine-based solutions. It is based on the idea that measurements can provide insights into the real world if transformed and analyzed over time. The model represents real-world objects and processes with simpler views and decomposable sub-processes. A case study demonstrates applying the model to process patient measurement data from a medical center over time and link it to medical events and expected outcomes. Future work involves developing a framework for both experts and non-experts to work with the domain models.
Valeri Labunets - Fast multiparametric wavelet transforms and packets for ima...AIST
This document discusses multiparametric wavelet transforms (WDT). It describes the structure of WDT, which is characterized by two sets of coefficients (h-coefficients and g-coefficients) related by a matrix equation. It also discusses arbitrary cyclic wavelet transforms (AWT) and their representation using Jacobi-Givens rotations and stairs-like structures. An example of an 8-level AWT is presented using these concepts.
Elena Bolshakova and Natalia Efremova - A Heuristic Strategy for Extracting T...AIST
The document describes a heuristic strategy for extracting terms from scientific texts. It discusses approaches to term extraction, including using statistical and linguistic criteria from large corpora or single texts. It also outlines developing term extraction procedures based on analyzing term types, structures, and contexts through linguistic patterns. The strategy was tested on Russian computer science and physics texts and compared to term dictionaries.
Alexandra Barysheva - Building Profiles of Blog Users Based on Comment Graph ...AIST
The document presents a method for building profiles of blog users based on analyzing comment graphs. The goal is to develop a language-independent tool to retrieve user profiles from online communities. The method studies user interactions in comment graphs, identifies attributes that can be retrieved from the graphs, and designs a profiling technique. It was tested on a dataset from Habrahabr involving over 2000 users. The results identified 5 types of user profiles based on clustering attributes like comments posted, received, and average distance in the graph. Further work could experiment on larger datasets and incorporate text from posts and comments.
Marina Danshina - The methodology of automated decryption of znamenny chantsAIST
1. The researchers created an automated system called "Computer Semiography" to decode Znamenny chants. The system consists of 5 modules: inputting chants into a database, reviewing manuscripts, forming linguistic and translation models, decoding chants, and a music editor.
2. The system can decode chants using either linear or Znamenny notation by applying rules from dictionaries and books. Researchers first build dictionaries from sources then use them to transform manuscripts into a linear notation.
3. The methodology allows producing three main components for decoding chants: a dictionary with translation rules, a translated manuscript version, and language and translation models. The work developed software to input, edit, and view chants in the database
Коберниченко Виктор Григорьевич, Сосновский Андрей Васильевич - Методы Оценив...AIST
Коберниченко Виктор Григорьевич, Сосновский Андрей Васильевич (Ural Federal University)
Методы Оценивания Интерферометрической Когерентности При Обработке Данных
AIST Conference 2015 http://aistconf.org
Thu Huong Nguyen - On Road Defects Detection and ClassificationAIST
This document discusses a method for detecting and classifying road defects using image processing and machine learning algorithms. The method involves collecting road pavement images, segmenting the images using graph cuts to identify regions of interest, extracting features from those regions, and classifying the defects using a random forest algorithm. Experimental results on road images from Irkutsk showed encouraging performance for automatic identification of cracks, potholes and other pavement defects.
The document summarizes information about the AIST 2014 conference, including that it received 74 submissions, mostly from Russian authors. Key highlights included the selection process, publication of accepted papers in Springer's Communications in Computer and Information Science series, and organization by a committee from Russia, UK, and other countries. It also lists sponsors and invited speakers for the conference.
Iosif Itkin - Network models for exchange trade analysisAIST
The document discusses software testing tools from Exactpro Systems for validating trading systems and ensuring data reconciliation. It introduces several tools the company offers: ClearTH for post-trade testing; MiniRobots for multi-threaded Java testing; Dolphin for market surveillance testing; Shsha for post-transactional analysis; Load Injector for load testing; and Sailfish for end-to-end testing. It also provides background on software quality assurance processes and examples of financial technology failures like the 2012 Knight Capital incident and issues with Facebook's NASDAQ IPO cross.
Artyom Makovetskii - An Efficient Algorithm for Total Variation DenoisingAIST
This document summarizes a research paper that analyzes the total variation denoising algorithm. It presents the following key points:
1. The total variation denoising model aims to minimize the sum of a fidelity term measuring noise and a regularization term measuring total variation.
2. The solution space can be reduced from bounded variation functions to piecewise constant functions on a given partition.
3. Explicit solutions are described for small values of the regularization parameter λ using Strong-Chan formulas, and these solutions are used to iteratively reduce the problem size and λ value.
4. The properties of extremal functions are proved, including uniqueness and behavior at discontinuity points depending on the sign of neighboring
Sergey Zaika and Andrew Toporkov - Semantic Web on Duty of E- Learning: Ontol...AIST
The document discusses an ontological approach for educating programmers using semantic web technologies. The authors developed an ontological model with 128 nodes and part-of relations. They also generated test questions using techniques like homonyms, quasi-synonyms, paronyms, and antonyms to differentiate meanings based on context. The authors thank the audience for their attention and provide their contact information.
Alexander Mikov - Program Tools for Dynamic Investigation of Social NetworksAIST
This document describes a simulation software tool called Triad.Net for investigating dynamic social networks. It allows modeling social networks and simulating information diffusion. Triad.Net includes components for model design, debugging, output analysis, security, and load balancing in distributed simulations. The software represents models using layers for structure, behavior, and messaging. It supports graph operations and standard network topologies. Triad.Net aims to help analyze hidden dependencies, structural properties, and conditions that impact simulation runs. Experiments show it can reduce rollback costs compared to optimistic simulation algorithms.
Andrey Kuznetsov and Vladislav Myasnikov - Using Efficient Linear Local Feat...AIST
The document proposes a new copy-move forgery detection algorithm using efficient linear local features. The key features of the algorithm are 100% recall, high calculation speed for real-time analysis, and 99.9% precision. It works by using a sliding window to analyze image fragments defined by a structural pattern. Hash values are calculated for each fragment based on linear local features and stored in a hash table. Duplicates are identified by finding hash values with a frequency greater than one. Experiments on satellite images show the algorithm has no false negatives and very low false positives, while being much faster than other methods.
Kaytoue Mehdi - Finding duplicate labels in behavioral data: an application f...AIST
The document discusses identifying duplicate labels, or aliases, in behavioral data from e-sports games. It presents a method to analyze confusion matrices from predictive models to identify pairs of labels that concentrate confusion, indicating they may belong to the same player using different aliases. The method extracts fuzzy concepts from the confusion matrix and scores candidate pairs based on their cosine similarity to rank and filter the most likely alias pairs. Experimental settings on real e-sports datasets are also discussed.
Quantum computing takes a giant leap forward from today’s technology—one that will forever alter our economic, industrial, academic, and societal landscape. This has massive implications for your customers in any industry including healthcare, energy, environmental systems, smart materials, and more. Learn how Microsoft is taking a unique revolutionary approach to quantum and how your customers can get started developing quantum solutions with the Quantum Development Kit.
The anatomy of a chemical reaction: Dissection by machine learning algorithmsAlex Clark
This document discusses using machine learning algorithms to analyze chemical reaction data. It describes how current reaction reporting formats are not well-suited for computational analysis. A more structured reporting format is proposed to fully describe reactions in a digitally friendly way, including specifying reactants, products, quantities, yields, and metrics like atom efficiency. This structured data would allow modeling of reaction substitutability and enable large-scale machine learning of chemical transformations.
An Introduction to Statistical Methods and Data Analysis.pdfSandra Valenzuela
This document provides a table of standard normal curve areas. The table lists z-scores from -5 to 5 in increments of 0.1 on the left side. For each z-score, the corresponding area under the standard normal curve is provided to four decimal places. The areas decrease as the z-scores decrease from 0 to -5 and increase as the z-scores increase from 0 to 5. Below the table, it notes that the areas were computed using the R function pnorm and that the shaded area represents the probability Pr(Z ≤ z).
Nataly Zhukova - Conceptual Model for Routine Measurements Analyses in Seman...AIST
The document presents a conceptual model for processing routine measurements in adaptive intelligent information systems. The model aims to help users solve domain-specific tasks by reducing large amounts of data, linking related pieces of information, and providing machine-based solutions. It is based on the idea that measurements can provide insights into the real world if transformed and analyzed over time. The model represents real-world objects and processes with simpler views and decomposable sub-processes. A case study demonstrates applying the model to process patient measurement data from a medical center over time and link it to medical events and expected outcomes. Future work involves developing a framework for both experts and non-experts to work with the domain models.
Valeri Labunets - Fast multiparametric wavelet transforms and packets for ima...AIST
This document discusses multiparametric wavelet transforms (WDT). It describes the structure of WDT, which is characterized by two sets of coefficients (h-coefficients and g-coefficients) related by a matrix equation. It also discusses arbitrary cyclic wavelet transforms (AWT) and their representation using Jacobi-Givens rotations and stairs-like structures. An example of an 8-level AWT is presented using these concepts.
Elena Bolshakova and Natalia Efremova - A Heuristic Strategy for Extracting T...AIST
The document describes a heuristic strategy for extracting terms from scientific texts. It discusses approaches to term extraction, including using statistical and linguistic criteria from large corpora or single texts. It also outlines developing term extraction procedures based on analyzing term types, structures, and contexts through linguistic patterns. The strategy was tested on Russian computer science and physics texts and compared to term dictionaries.
Alexandra Barysheva - Building Profiles of Blog Users Based on Comment Graph ...AIST
The document presents a method for building profiles of blog users based on analyzing comment graphs. The goal is to develop a language-independent tool to retrieve user profiles from online communities. The method studies user interactions in comment graphs, identifies attributes that can be retrieved from the graphs, and designs a profiling technique. It was tested on a dataset from Habrahabr involving over 2000 users. The results identified 5 types of user profiles based on clustering attributes like comments posted, received, and average distance in the graph. Further work could experiment on larger datasets and incorporate text from posts and comments.
Marina Danshina - The methodology of automated decryption of znamenny chantsAIST
1. The researchers created an automated system called "Computer Semiography" to decode Znamenny chants. The system consists of 5 modules: inputting chants into a database, reviewing manuscripts, forming linguistic and translation models, decoding chants, and a music editor.
2. The system can decode chants using either linear or Znamenny notation by applying rules from dictionaries and books. Researchers first build dictionaries from sources then use them to transform manuscripts into a linear notation.
3. The methodology allows producing three main components for decoding chants: a dictionary with translation rules, a translated manuscript version, and language and translation models. The work developed software to input, edit, and view chants in the database
Коберниченко Виктор Григорьевич, Сосновский Андрей Васильевич - Методы Оценив...AIST
Коберниченко Виктор Григорьевич, Сосновский Андрей Васильевич (Ural Federal University)
Методы Оценивания Интерферометрической Когерентности При Обработке Данных
AIST Conference 2015 http://aistconf.org
Thu Huong Nguyen - On Road Defects Detection and ClassificationAIST
This document discusses a method for detecting and classifying road defects using image processing and machine learning algorithms. The method involves collecting road pavement images, segmenting the images using graph cuts to identify regions of interest, extracting features from those regions, and classifying the defects using a random forest algorithm. Experimental results on road images from Irkutsk showed encouraging performance for automatic identification of cracks, potholes and other pavement defects.
The document summarizes information about the AIST 2014 conference, including that it received 74 submissions, mostly from Russian authors. Key highlights included the selection process, publication of accepted papers in Springer's Communications in Computer and Information Science series, and organization by a committee from Russia, UK, and other countries. It also lists sponsors and invited speakers for the conference.
Iosif Itkin - Network models for exchange trade analysisAIST
The document discusses software testing tools from Exactpro Systems for validating trading systems and ensuring data reconciliation. It introduces several tools the company offers: ClearTH for post-trade testing; MiniRobots for multi-threaded Java testing; Dolphin for market surveillance testing; Shsha for post-transactional analysis; Load Injector for load testing; and Sailfish for end-to-end testing. It also provides background on software quality assurance processes and examples of financial technology failures like the 2012 Knight Capital incident and issues with Facebook's NASDAQ IPO cross.
Artyom Makovetskii - An Efficient Algorithm for Total Variation DenoisingAIST
This document summarizes a research paper that analyzes the total variation denoising algorithm. It presents the following key points:
1. The total variation denoising model aims to minimize the sum of a fidelity term measuring noise and a regularization term measuring total variation.
2. The solution space can be reduced from bounded variation functions to piecewise constant functions on a given partition.
3. Explicit solutions are described for small values of the regularization parameter λ using Strong-Chan formulas, and these solutions are used to iteratively reduce the problem size and λ value.
4. The properties of extremal functions are proved, including uniqueness and behavior at discontinuity points depending on the sign of neighboring
Sergey Zaika and Andrew Toporkov - Semantic Web on Duty of E- Learning: Ontol...AIST
The document discusses an ontological approach for educating programmers using semantic web technologies. The authors developed an ontological model with 128 nodes and part-of relations. They also generated test questions using techniques like homonyms, quasi-synonyms, paronyms, and antonyms to differentiate meanings based on context. The authors thank the audience for their attention and provide their contact information.
Alexander Mikov - Program Tools for Dynamic Investigation of Social NetworksAIST
This document describes a simulation software tool called Triad.Net for investigating dynamic social networks. It allows modeling social networks and simulating information diffusion. Triad.Net includes components for model design, debugging, output analysis, security, and load balancing in distributed simulations. The software represents models using layers for structure, behavior, and messaging. It supports graph operations and standard network topologies. Triad.Net aims to help analyze hidden dependencies, structural properties, and conditions that impact simulation runs. Experiments show it can reduce rollback costs compared to optimistic simulation algorithms.
Andrey Kuznetsov and Vladislav Myasnikov - Using Efficient Linear Local Feat...AIST
The document proposes a new copy-move forgery detection algorithm using efficient linear local features. The key features of the algorithm are 100% recall, high calculation speed for real-time analysis, and 99.9% precision. It works by using a sliding window to analyze image fragments defined by a structural pattern. Hash values are calculated for each fragment based on linear local features and stored in a hash table. Duplicates are identified by finding hash values with a frequency greater than one. Experiments on satellite images show the algorithm has no false negatives and very low false positives, while being much faster than other methods.
Kaytoue Mehdi - Finding duplicate labels in behavioral data: an application f...AIST
The document discusses identifying duplicate labels, or aliases, in behavioral data from e-sports games. It presents a method to analyze confusion matrices from predictive models to identify pairs of labels that concentrate confusion, indicating they may belong to the same player using different aliases. The method extracts fuzzy concepts from the confusion matrix and scores candidate pairs based on their cosine similarity to rank and filter the most likely alias pairs. Experimental settings on real e-sports datasets are also discussed.
Quantum computing takes a giant leap forward from today’s technology—one that will forever alter our economic, industrial, academic, and societal landscape. This has massive implications for your customers in any industry including healthcare, energy, environmental systems, smart materials, and more. Learn how Microsoft is taking a unique revolutionary approach to quantum and how your customers can get started developing quantum solutions with the Quantum Development Kit.
The anatomy of a chemical reaction: Dissection by machine learning algorithmsAlex Clark
This document discusses using machine learning algorithms to analyze chemical reaction data. It describes how current reaction reporting formats are not well-suited for computational analysis. A more structured reporting format is proposed to fully describe reactions in a digitally friendly way, including specifying reactants, products, quantities, yields, and metrics like atom efficiency. This structured data would allow modeling of reaction substitutability and enable large-scale machine learning of chemical transformations.
An Introduction to Statistical Methods and Data Analysis.pdfSandra Valenzuela
This document provides a table of standard normal curve areas. The table lists z-scores from -5 to 5 in increments of 0.1 on the left side. For each z-score, the corresponding area under the standard normal curve is provided to four decimal places. The areas decrease as the z-scores decrease from 0 to -5 and increase as the z-scores increase from 0 to 5. Below the table, it notes that the areas were computed using the R function pnorm and that the shaded area represents the probability Pr(Z ≤ z).
This document discusses high-throughput screening (HTS) workflows for identifying biologically active small molecules. It describes how robots are used to rapidly screen large libraries of compounds in assays and generate large datasets. Statistical and machine learning methods in R can then be used to build predictive models from these datasets to identify promising leads and guide the screening of additional compounds. Caveats regarding the applicability of models to new chemical spaces are also discussed.
Linear regression an 80 year study of the dow jones industrial averageTehyaSingleton
Linear regression was used to model the relationship between the Dow Jones Industrial Average (DJIA) price and years since 1930 over an 80 year period. The results showed a strong positive linear relationship where DJIA price increases by about 125 points for each additional year. The slope of the linear model indicates that DJIA price rises as years since 1930 increases. The y-intercept of the model, which is the hypothetical DJIA price at year 0 (1930), provides meaningful context about the starting price over the 80 years analyzed.
Linear regression an 80 year study of the dow jones industrial averageTehyaSingleton
Linear regression was used to model the relationship between the Dow Jones Industrial Average (DJIA) price and years since 1930 over an 80 year period. The results showed a strong positive linear relationship where DJIA price increases by about 125 points for each additional year. The regression equation determined that DJIA price equals 125.3 times the number of years since 1930 minus 2.4425. While DJIA price has generally increased over the eight decades, the model suggests it would have been negative in 1930 based on the y-intercept value.
Detecting Malicious Websites using Machine LearningAndrew Beard
We present a set of newly tuned algorithms that can distinguish between malicious and non-malicious websites with a high degree of accuracy using Machine Learning (ML). We use the Bro IDS/IPS tool for extracting the SSL certificates from network traffic and training the ML algorithms.
The extracted SSL attributes are then loaded into multiple ML frameworks such as Splunk, AWS ML and we run a series of classification algorithms to identify those attributes that correlate with malicious sites.
Our analysis shows that there are a number of emerging patterns that even allow for identification of high-jacked devices and self-signed certificates. We present the results of our analysis which show which attributes are the most relevant for detecting malicious SSL certificates and as well the performance of the ML algorithms.
The document appears to be grades from a university course on reading and writing. It lists the student IDs and grades received on various assignments, activities, and assessments conducted throughout the course including exams, workshops, discussions, and research activities. The assignments were completed using online tools and platforms like Dropbox, Turnitin, blogs, and forums.
1) The document discusses models for estimating car trip generation in Nairobi and Dar-es-Salaam. It estimates four types of models: with and without car ownership as an explanatory variable, two-stage models, and joint car ownership and trip generation models.
2) The results show household income, number of workers and drivers, and car ownership positively influence trip generation. However, some differences exist between the cities.
3) Models 3 and 4, which account for potential endogeneity between car ownership and trips, better explain trip generation in both cities compared to Models 1 and 2.
1) The document discusses models for estimating car trip generation in Nairobi and Dar-es-Salaam. It estimates four types of models: with and without car ownership as an explanatory variable, two-stage models, and joint car ownership and trip generation models.
2) The results show household income, number of workers and drivers, and car ownership positively influence trip generation. However, some differences exist between the cities.
3) Models 3 and 4, which account for potential endogeneity between car ownership and trips, better explain trip generation in both cities compared to Models 1 and 2.
Cost of Living, Linear Regression, Multiple Linear RegressionFaisal Akbar
The document discusses various types of regression analysis used to analyze the relationship between variables. Simple linear regression examines the relationship between one independent and one dependent variable. Multiple linear regression examines the relationship between one dependent and multiple independent variables. Rent was found to have the strongest correlation and best predict overall cost of living, while newspaper had the weakest correlation. A multiple linear regression model found rent, public transport, CD, news, coffee, and fast food costs together can predict 89.2% of the variation in overall cost of living.
Rent has the strongest positive correlation with overall cost of living, making it the best predictor. Newspaper costs have the strongest negative correlation, providing an unusual relationship. A multiple regression analysis found rent, public transport, compact disc and fast food costs significantly predict cost of living, explaining around 89% of the variation. Coffee has the weakest relationship and is the worst individual predictor of overall living expenses.
How do you cut the Big Data clutter and tell interesting, insightful and impacting stories? This session talks about the need for Data Visualization & how Visual stories can come to the aid of the Big Data problem associated with meaningful consumption. The point is illustrated by leveraging several industry case studies.
Tapping into the value on a transport corridorTristan Wiggill
A presentation by Nic Cloete-Hopkins, Transnet Senior Lecturer in Systems Engineering at Wits University at the Transport Forum special interest group in collaboration with MCLI in Mbombela on 4 February 2016.
The theme for the event was: "Transport Corridors". The topic of the presentation was: "Tapping into the value on a Transport Corridor".
More like this on www.transportworldafrica.co.za
Classification of Human's Driving Behavior using Support Vector MachineKitsukawa Yuki
This document describes a study that used support vector machines to classify human driving behavior based on surrounding environment and conditions. The study analyzed a dataset containing vehicle sensor and location data collected at one-second intervals. It classified whether a driver braked based on factors like velocity and pedestrians nearby, and whether pedestrians were present based on driving actions like braking. The best performing model classified braking behavior accurately 88.3% of the time and pedestrian presence 88.3% of the time. Future work could improve the model through additional data factors and collection of more data.
The document discusses computer vision and machine vision. It explains that computer vision extracts scene information from images and video, such as geometry and objects. Machine vision is used in engineered environments with special markers or cameras. Basic useful information that can be obtained from a camera includes the location of markers and object segmentation. Commercial tracking systems use infrared or visible light and cameras to track markers.
Similar to Dmitry Berg, Olga Zvereva - Identification Of Autopoietic Communication Patterns In Social And Economic Networks (20)
Alexey Mikhaylichenko - Automatic Detection of Bone Contours in X-Ray ImagesAIST
This document summarizes an algorithm for automatically segmenting and detecting joints in x-ray images. The algorithm involves several steps: (1) computing an edge map of the image, (2) determining binarization thresholds, (3) thinning edges, (4) binarizing the image, and (5) chaining edges together. The algorithm is compared to the Canny edge detector and is shown to achieve a 74% success rate on joint images. Processing time is improved by implementing a multi-threaded version. Challenges include false edge detection and discontinuities between edge fragments.
Евгений Цымбалов, Webgames - Методы машинного обучения для задач игровой анал...AIST
ML is helping a large Russian game developer and publisher called WebGames analyze data from their free-to-play games. They collect over 80 million records daily from their 400k daily players across various platforms. They use ML for tasks like churn prediction, revenue prediction, user classification, A/B testing, balance, and recommendations. Specifically, they build 30 different models to predict LTV for users based on their behavior in the first 30 days. They also use kNN and cohort-based approaches for user classification and Bayesian A/B testing to dynamically adjust testing over time. Rule-based modeling and midgame support based on classification help balance games. Content recommendations are done through static and dynamic clustering.
Александр Москвичев, EveResearch - Алгоритмы анализа данных в маркетинговых и...AIST
The document discusses various economic models related to utility and consumer choice. It introduces concepts like cardinal and ordinal utility and discusses modeling utility as a probability of success. It also discusses simple models of consumer choice that involve choosing a product based on type and brand or type and price. Additionally, it discusses models where neighbors' choices or prices can influence individual choices and mentions discounts and bundles. Graphs are shown comparing product turnover before and after implementing a neighbors effect model.
1) Exactpro is a specialist QA firm focused on testing financial systems that was acquired by the London Stock Exchange Group in 2015.
2) The London Stock Exchange Group is a leading international exchange group that traces its history back to 1698 and has over 5,500 employees.
3) Exactpro uses automated testing tools like Sailfish and ClearTH to test systems, as well as techniques like formal verification, crowd-sourced testing, and machine learning.
Nikolay Karpov - Evolvable Semantic Platform for Facilitating Knowledge ExchangeAIST
This document describes an evolvable semantic platform called EXPERTIZE that was developed to facilitate knowledge exchange between experts at a university. EXPERTIZE analyzes unstructured text from news and matches it to the skills of university experts, as defined in their personal ontologies, in order to recommend relevant experts. It uses a latent Dirichlet allocation algorithm to perform the semantic matching. The system was implemented and evaluated, showing its ability to successfully recommend experts and categories for news items.
George Moiseev - Classification of E-commerce Websites by Product CategoriesAIST
This document describes a study that classified e-commerce websites by the products they sell. It discusses preprocessing web pages, extracting features using TF-IDF with additional weighting for tags, and classifying pages using a support vector machine. The results show that considering information from other pages in addition to the main page improved classification accuracy, with an average F-score of 0.81 for product type classification when using all page information with the tag weighting.
Elena Bruches - The Hybrid Approach to Part-of-Speech DisambiguationAIST
The document describes a hybrid approach to part-of-speech disambiguation that combines neural networks and manually crafted rules. The algorithm uses neural networks to generate a set of possible part-of-speech tags for each word, and rule-based tagging to generate another set. The final set of tags is the intersection of these two sets, or their union if the intersection is empty. The approach achieved 96.11% precision on one corpus and 86.39% precision on another larger corpus.
Edward Klyshinsky - The Corpus of Syntactic Co-occurences: the First GlanceAIST
The document discusses the Corpus of Syntactic Co-occurrences, which aims to provide a corpus for students learning Russian that contains correct word combinations. It notes existing corpora like the Russian National Corpus are too large and technical for beginners. The CoSyCo extracts unambiguous syntactic phrases from Russian texts that could help learners. It uses various news and technical texts totaling over 15 billion words. Examples of extracted phrases are provided. The CoSyCo site is mentioned and future plans outlined, such as enlarging the phrase list, filtering repeats and strange combinations, improving the design, and making the co-occurrence database clearer.
Galina Lavrentyeva - Anti-spoofing Methods for Automatic Speaker Verification...AIST
This document discusses anti-spoofing methods for automatic speaker verification systems. It summarizes various spoofing methods like replay attacks, voice conversion, and text-to-speech synthesis that attempt to manipulate biometric systems. The document then outlines the ASVspoof 2015 challenge on spoofing detection and the system submitted by the authors that achieved 2nd place. It details the authors' system including front-end preprocessing, feature extraction using magnitude, phase and high-level features, and back-end classifiers like SVM and neural networks. The system fused multiple feature types to achieve robust spoofing detection.
Oleksandr Frei and Murat Apishev - Parallel Non-blocking Deterministic Algori...AIST
This document describes parallel algorithms for topic modeling, including synchronous, asynchronous, and deterministic asynchronous algorithms. The synchronous offline algorithm splits the document collection into batches and has each thread process one batch at a time. The asynchronous online algorithm has processor threads process batches concurrently while a merger thread accumulates and merges results to recalculate model parameters. To make the algorithm deterministic, the deterministic asynchronous approach has each thread process batches and write results directly without a merger thread.
Valeri Labunets - The bichromatic excitable Schrodinger metamediumAIST
This document describes research into modeling wave phenomena like particle motion and interference using a cellular automata approach called an excitable metamedium. It can simulate the Schrodinger equation by representing diffusion as complex numbers across cells. The researchers extended this to use triplet "color" numbers for diffusion coefficients, allowing visualization of properties like hue, saturation and lightness. Experiments demonstrated particle motion, interference and blending effects using different color diffusion values. Unusual geometries were also explored by changing the definition of the imaginary unit, affecting the behavior of color wave propagation in interesting ways.
Alexander Karkishchenko - Threefold Symmetry Detection in Hexagonal Images Ba...AIST
This document discusses a method for detecting threefold symmetry in hexagonal images using finite Eisenstein fields. It begins with an introduction to symmetry detection and issues that arise when applying existing continuous techniques to digital images. It then describes finite Eisenstein fields, which are constructed as finite fields analogous to the complex integers. Elements of these fields correspond naturally to hexagons, allowing hexagonal images to be represented as functions over the fields. Polar coordinate transformations are introduced to represent field elements in exponential form, enabling the transfer of continuous symmetry detection methods to digital hexagonal images. In summary, the document proposes a novel approach for symmetry detection in hexagonal images based on the algebraic structure of finite Eisenstein fields.
Olesia Kushnir - Reflection Symmetry of Shapes Based on Skeleton Primitive Ch...AIST
1) The document proposes a method for detecting approximate reflection symmetry in shapes based on representing the skeleton of the shape as a chain of primitives.
2) It describes representing the skeleton as a chain of primitives containing the length and angle of each skeleton edge. This chain can then be divided into two sub-chains and aligned to evaluate symmetry.
3) The algorithm involves choosing start and end points on the skeleton to divide it into left and right sub-chains, adjusting the right sub-chain to be reflected, and calculating the dissimilarity between the sub-chains to determine the symmetry.
Oxana Logunova - The Results Of Sulfur Print Image Classification Of Section ...AIST
This presentation discusses the results of classifying sulfur print images of steel billet sections into three classes (A, B, C) using fuzzy logic. Membership functions were formed for the linguistic variables that describe image characteristics like brightness threshold, maximum brightness on each side of the threshold. Decision rules were developed to classify images when characteristics fall into ambiguous overlapping regions between different classes. The presentation concludes with thanks for attention.
Anton Korsakov - Determination of an unmanned mobile object orientation by na...AIST
1) The document presents an algorithm for determining the orientation of an unmanned mobile object using natural landmarks.
2) The algorithm uses a classifier to detect three natural landmarks in video from the mobile object. It then calculates the object's orientation based on the angles between the landmarks.
3) Experimental tests found the algorithm could determine orientation accurately to within 1.6 degrees on average using this landmark-based approach.
Anton Korsakov - Determination of an unmanned mobile object orientation by na...
Dmitry Berg, Olga Zvereva - Identification Of Autopoietic Communication Patterns In Social And Economic Networks
1. IDENTIFICATION OF AUTOPOIETIC
COMMUNICATION PATTERNS IN SOCIAL
AND ECONOMIC NETWORKS
The 4th international conference on Analysis of Images, Social
Networks, and Texts, April, 2015, Ekaterinburg
2. • Communications develop the basis for social and
economic system functioning
• A set of economic agent communications forms a
network
• One of the most important system characteristic is its
ability to reproduce itself (autopoiesis)
• Autopoiesis is realized according to Maturana and
Varella’s theory by cycled communications in the closed
contour.
• We tried to find such contours (closed paths) in
economic and social networks and considered them to
form autopoietic patterns in the networks under
discussion.
3. 1. RESEARCH GOALS
2. MUNICIPAL ECONOMIC NETWORK
1. CONSUMPTION MATRIX
2. NETWORK DIGRAPH, NETWORK INDICES
3. REVEALING OF THE CLOSED PATHS
4. BALANCED CONSUMPTION MATRIX
3. AGENT-BASED MODEL
4. SOCIAL NETWORK
1. NETWORK SOCIOGRAM
2. NETWORK INDICES
5. HYPOTHESES
4. • Niklas Luhmann. Social systems. Sketch of the general theory. St. Petersburg:
Science. 2007. 668p
• Manuel Castells. The Rise of the Network Society, The Information Age:
Economy, Society and Culture, Vol. I. Cambridge, MA; Oxford, UK: Blackwell
(1996) (second edition, 2000), 656 р.
• H.R.Maturana, F.J.Varela Autopoiesis and Cognition: The Realization of the
Living (Boston Studies in the Philosophy of Science, Vol. 42) Boston: D.
Reidel Publishing Company; 1st edition 1980. 171 p
• Leontief V.V. Essays in Economics. Theories, Theorizing, Facts and Policies.
Moscow: Politizdat. 1990. 415p.
• L. da F.Costa, F.A. Rodrigues, G. Travieso, P.R. Villas Boas. Characterization
of Complex Networks: A Survey of Measurements. Advances in Physics, 56:1
(2007), pp. 167-242.
• Tore Opsahl, Filip Agneessens, John Skvoretz. Node centrality in weighted
networks: Generalizing degree and shortest paths. Social Networks 32 (2010),
pp. 245–251.
5. • To observe communications of real economic agents and to receive
a formal matrix representation of these communications
• To explore the results of communications in the social network
• Using SNA methods and instruments to find the autopoietic contours in the
social network
• To prove with the help of agent-based model that the balance (Leontieff’s
equilibrium) exists
• Using SNA methods and instruments to find the autopoietic contours
(closed paths) and make the matrix balanced
• To receive the main network characteristics
6. Enterprise list )
1. agricultural farm (cereal and industrial crops cultivating);
2. meat and dairy farm (meat and dairy production);
3. poultry farm (poultry breeding and egg production);
4. meat processing plant (production of semi-finished goods from meat and
poultry);
5. dairy plant (production of dairy goods);
6. bakery (bakery product manufacturing);
7. flour mill (flour production);
8. feed mill (feed production);
9. furniture factory (home and office furniture production);
10. autoservice workshop (car maintenance and repair);
11. trucking company (shipping operations).
9. № Contour Circularity Link Weights Autopoiesis
1 1→8→2→1 yes 1340680+419318+405 yes
2 1→7→6→12→1 yes 496000+4081+27514+
14000
yes
3 1→8→3→12→1 yes 1340680+419318+8227+
14000
yes
4 2→4→12→2 yes 53993+109460+17000 yes
5 2→5→12→2 yes 590322+79389+17000 yes
6 3→12→8→3 yes 8227+30000+419318 yes
7 4→12→3→4 yes 109460+40450+79100 yes
8 6→12→7→6 yes 27514+16000+4081 yes
9 11→12→10→11 yes 16280+7000+33 yes
10 12→1→8←12 no 14000+1340680+30000 no
11 12→2→1←12 no 17000+405+14000 no
12 8→2→4→12→3←8 no 8726+53993+109460+40450+
419318
no
10. • Population (12) is the central node (a star) in the
network which integrates separate autopoietic contours
in one autopoietic pattern
• This node has the maximum centrality values in the
network:
– Centrality in-degree is 0,636
– Centrality out-degree is 1,0 (connected with all other
agents).
– The value of Freeman’s betweeness centrality is
equal to 88,833 (more than 88% of the shortest paths
from one agent to another in this network includes
the 12th agent).
14. № Contour Circularity Weights (marks) Autopoies
is
1 3→11→4→5→3 yes 10+3+4+2=19 yes
2 3→11→6→5→3 yes 10+3+5+3=21 yes
3 3→11→7→5→3 yes 10+4+41+3=57 yes
4 3→11→12→5→3 yes 10+4+75+3=92 yes
5 3→11→4→9→3 yes 10+3+4-5 no
6 3→11→6→9→3 yes 10+3+5-5 no
7 3→11→7→9→3 yes 10+3+50-5 no
8 3→11→12→9→3 yes 10+3+58-5 no
9 3←4→5→3 no 4+4+2=10 no
10 3←6→7→9→3 no 5+5+50+9=19 no
11 3←7→11→4→3 no 9+16+3+4=32 no
15. • This network is not highly centralized (the
Network Centralization Index is only 6,83% in
comparison to 77,19% in the economic
network).
• As considered to the agents who integrate these
contours in one pattern, they are the 9th and
11th ones. As was estimated with the help of
UCINET 6.0 the 9th and 11th agents have the
maximum values of Freeman’s centrality
betweeness (11,943 and 10,167 respectively) in
this network.
16. • There are autopoietic patterns in economic and social networks
• Every pattern is a set of closed circular contours with some corporate elements
(nodes/agents). As usual, these “corporate” agents have high values of centrality
measures.
• Consumption matrix of an autopoietic pattern is the balanced one and reproduction
in this case can be described with the help of Leontief’s intersectoral equilibrium.
• Two technologies are rather useful for social and economic system analysis: SNA for
structural properties revealing and agent-based – for dynamic characteristics
determining