Graphical data aggregation and analysis in dedicated mobile device networks.
The problem of inefficient processing in the Big Data industry is touched upon. A detailed analysis of the various means to increase the percentage of processed data is provided and the experimental implementation of a way to obtain and preprocess data in a mobile device network in real-time mode is...
Збережено в:
| Дата: | 2019 |
|---|---|
| Автори: | , |
| Формат: | Стаття |
| Мова: | Ukrainian |
| Опубліковано: |
Інститут проблем реєстрації інформації НАН України
2019
|
| Теми: | |
| Онлайн доступ: | http://drsp.ipri.kiev.ua/article/view/180137 |
| Теги: |
Додати тег
Немає тегів, Будьте першим, хто поставить тег для цього запису!
|
| Назва журналу: | Data Recording, Storage & Processing |
Репозитарії
Data Recording, Storage & Processing| id |
drspiprikievua-article-180137 |
|---|---|
| record_format |
ojs |
| institution |
Data Recording, Storage & Processing |
| collection |
OJS |
| language |
Ukrainian |
| topic |
концепт Big Data технологія Deep Learning технологія GPGPU операційна система Android бібліотека TensorFlow бібліотека CNTK бібліотека PyTorch бібліотека MxNet бібліотека Caffe нейронна мережа Mobile Net нейронна мережа Squeeze Net Big Data concept Deep Learning technology GPGPU technology Android operation system TensorFlow framework CNTK framework PyTorch framework MxNet framework Caffe framework Mobile Net neuron network Squeeze Net neuron network Inception Net neuron network NAS Net neuron network |
| spellingShingle |
концепт Big Data технологія Deep Learning технологія GPGPU операційна система Android бібліотека TensorFlow бібліотека CNTK бібліотека PyTorch бібліотека MxNet бібліотека Caffe нейронна мережа Mobile Net нейронна мережа Squeeze Net Big Data concept Deep Learning technology GPGPU technology Android operation system TensorFlow framework CNTK framework PyTorch framework MxNet framework Caffe framework Mobile Net neuron network Squeeze Net neuron network Inception Net neuron network NAS Net neuron network Pogorilyy, S. D. Chechula, M. V. Graphical data aggregation and analysis in dedicated mobile device networks. |
| topic_facet |
концепт Big Data технологія Deep Learning технологія GPGPU операційна система Android бібліотека TensorFlow бібліотека CNTK бібліотека PyTorch бібліотека MxNet бібліотека Caffe нейронна мережа Mobile Net нейронна мережа Squeeze Net Big Data concept Deep Learning technology GPGPU technology Android operation system TensorFlow framework CNTK framework PyTorch framework MxNet framework Caffe framework Mobile Net neuron network Squeeze Net neuron network Inception Net neuron network NAS Net neuron network |
| format |
Article |
| author |
Pogorilyy, S. D. Chechula, M. V. |
| author_facet |
Pogorilyy, S. D. Chechula, M. V. |
| author_sort |
Pogorilyy, S. D. |
| title |
Graphical data aggregation and analysis in dedicated mobile device networks. |
| title_short |
Graphical data aggregation and analysis in dedicated mobile device networks. |
| title_full |
Graphical data aggregation and analysis in dedicated mobile device networks. |
| title_fullStr |
Graphical data aggregation and analysis in dedicated mobile device networks. |
| title_full_unstemmed |
Graphical data aggregation and analysis in dedicated mobile device networks. |
| title_sort |
graphical data aggregation and analysis in dedicated mobile device networks. |
| title_alt |
Агрегація та аналіз графічних даних у розподіленій мережі мобільних пристроїв |
| description |
The problem of inefficient processing in the Big Data industry is touched upon. A detailed analysis of the various means to increase the percentage of processed data is provided and the experimental implementation of a way to obtain and preprocess data in a mobile device network in real-time mode is shown.During the analysis of the subject, the next fields of research were observed: Deep Learning, Machine Learning, Big Data, and GPGPU technology. The increasing numbers of publications, especially of papers that include mobile networks, were emphasized as evidence of rapid growth of the industry and a transition of neuron network algorithms toward mobile operating systems.Further analysis was focused on highlighting the most relevant and perspective objects for the research. The analysis showed that amidst currently most innovative and broadly widespread operation systems and frameworks to implement and engage neuron network algorithms Android operation system and TensorFlow framework has the most significant advantages.Due to the purpose of developing an experimental solution based on mobile device network and neuron network, different classes and types of neuron network architectures were explored. Two major types of mobile neuron networks such as quantized and integer neuron networks and the principal dissimilarity between them were described. Various neuron networks were tested on mobile devices with Internet connection via specially developed auxiliary software using GPGPU technology. Experimental results had shown that modern smartphones such as Huawei P20-Pro are capable to analyze, store and transmit the incoming from its camera sensor information at a rate of up to 40 frames per second. The usage of mobile GPU for improving the performance of the neuron networks was proved to be effective as such the number of frames processed by a neuron network per second can be elevated up to 10 times.The experimental software that task was to search for the given by user object in a real-time mode using mobile device network as both — server and processing nodes — proved to be a violable solution for the increasing amount of preprocessed up-to-the-minute information in the Big Data industry after the diligent research.As a summary, it needs to be stated that current progress in mobile device industry is making possible to bring neuron network technologies on mobile platforms and expand the capability of data aggregating services through the use of modern Deep Learning frameworks and GPGPU technology. |
| publisher |
Інститут проблем реєстрації інформації НАН України |
| publishDate |
2019 |
| url |
http://drsp.ipri.kiev.ua/article/view/180137 |
| work_keys_str_mv |
AT pogorilyysd graphicaldataaggregationandanalysisindedicatedmobiledevicenetworks AT chechulamv graphicaldataaggregationandanalysisindedicatedmobiledevicenetworks AT pogorilyysd agregacíâtaanalízgrafíčnihdanihurozpodíleníjmerežímobílʹnihpristroív AT chechulamv agregacíâtaanalízgrafíčnihdanihurozpodíleníjmerežímobílʹnihpristroív |
| first_indexed |
2024-04-21T19:34:05Z |
| last_indexed |
2024-04-21T19:34:05Z |
| _version_ |
1796974099609681920 |
| spelling |
drspiprikievua-article-1801372019-12-10T11:25:26Z Graphical data aggregation and analysis in dedicated mobile device networks. Агрегація та аналіз графічних даних у розподіленій мережі мобільних пристроїв Pogorilyy, S. D. Chechula, M. V. концепт Big Data технологія Deep Learning технологія GPGPU операційна система Android бібліотека TensorFlow бібліотека CNTK бібліотека PyTorch бібліотека MxNet бібліотека Caffe нейронна мережа Mobile Net нейронна мережа Squeeze Net Big Data concept Deep Learning technology GPGPU technology Android operation system TensorFlow framework CNTK framework PyTorch framework MxNet framework Caffe framework Mobile Net neuron network Squeeze Net neuron network Inception Net neuron network NAS Net neuron network The problem of inefficient processing in the Big Data industry is touched upon. A detailed analysis of the various means to increase the percentage of processed data is provided and the experimental implementation of a way to obtain and preprocess data in a mobile device network in real-time mode is shown.During the analysis of the subject, the next fields of research were observed: Deep Learning, Machine Learning, Big Data, and GPGPU technology. The increasing numbers of publications, especially of papers that include mobile networks, were emphasized as evidence of rapid growth of the industry and a transition of neuron network algorithms toward mobile operating systems.Further analysis was focused on highlighting the most relevant and perspective objects for the research. The analysis showed that amidst currently most innovative and broadly widespread operation systems and frameworks to implement and engage neuron network algorithms Android operation system and TensorFlow framework has the most significant advantages.Due to the purpose of developing an experimental solution based on mobile device network and neuron network, different classes and types of neuron network architectures were explored. Two major types of mobile neuron networks such as quantized and integer neuron networks and the principal dissimilarity between them were described. Various neuron networks were tested on mobile devices with Internet connection via specially developed auxiliary software using GPGPU technology. Experimental results had shown that modern smartphones such as Huawei P20-Pro are capable to analyze, store and transmit the incoming from its camera sensor information at a rate of up to 40 frames per second. The usage of mobile GPU for improving the performance of the neuron networks was proved to be effective as such the number of frames processed by a neuron network per second can be elevated up to 10 times.The experimental software that task was to search for the given by user object in a real-time mode using mobile device network as both — server and processing nodes — proved to be a violable solution for the increasing amount of preprocessed up-to-the-minute information in the Big Data industry after the diligent research.As a summary, it needs to be stated that current progress in mobile device industry is making possible to bring neuron network technologies on mobile platforms and expand the capability of data aggregating services through the use of modern Deep Learning frameworks and GPGPU technology. Проаналізовано напрямки досліджень у галузі концептів Big Data, розподілених мереж мобільних пристроїв і Deep Learning. Кількісно охарактеризовано та порівняно інтенсивність розвитку сучасних бібліотек нейронних мереж для використання технологій Deep Learning, Big Data, GPGPU. Розроблено застосування для дослідження роботи згорткових нейронних мереж різної архітектури із використанням потуж-ностей мобільних CPU та GPU і з використанням API для нейронних мереж на операційній системі Android. Розроблено застосування для агрегації проаналізованих у реальному часі даних, структуризації даних на сервері та досліджено роботу застосування в мережах WiFi, 3G та 4G. Проведено аналіз різних шляхів агрегації даних. Інститут проблем реєстрації інформації НАН України 2019-11-21 Article Article application/pdf http://drsp.ipri.kiev.ua/article/view/180137 10.35681/1560-9189.2019.21.2.180137 Data Recording, Storage & Processing; Vol. 21 No. 2 (2019); 21-33 Регистрация, хранение и обработка данных; Том 21 № 2 (2019); 21-33 Реєстрація, зберігання і обробка даних; Том 21 № 2 (2019); 21-33 1560-9189 uk http://drsp.ipri.kiev.ua/article/view/180137/184144 Авторське право (c) 2021 Реєстрація, зберігання і обробка даних |