Bulletin of the Polish Academy of Sciences-Technical Sciences最新文献_第2页

Theory II: Deep learning and optimization 理论二:深度学习和优化

IF 1.2 4区工程技术

Bulletin of the Polish Academy of Sciences-Technical Sciences Pub Date : 2023-11-06 DOI: 10.24425/BPAS.2018.125925

T. Poggio, Q. Liao

{"title":"Theory II: Deep learning and optimization","authors":"T. Poggio, Q. Liao","doi":"10.24425/BPAS.2018.125925","DOIUrl":"https://doi.org/10.24425/BPAS.2018.125925","url":null,"abstract":"Bull. Pol. Ac.: Tech. 66(6) 2018 Abstract. The landscape of the empirical risk of overparametrized deep convolutional neural networks (DCNNs) is characterized with a mix of theory and experiments. In part A we show the existence of a large number of global minimizers with zero empirical error (modulo inconsistent equations). The argument which relies on the use of Bezout theorem is rigorous when the RELUs are replaced by a polynomial nonlinearity. We show with simulations that the corresponding polynomial network is indistinguishable from the RELU network. According to Bezout theorem, the global minimizers are degenerate unlike the local minima which in general should be non-degenerate. Further we experimentally analyzed and visualized the landscape of empirical risk of DCNNs on CIFAR-10 dataset. Based on above theoretical and experimental observations, we propose a simple model of the landscape of empirical risk. In part B, we characterize the optimization properties of stochastic gradient descent applied to deep networks. The main claim here consists of theoretical and experimental evidence for the following property of SGD: SGD concentrates in probability – like the classical Langevin equation – on large volume, ”flat” minima, selecting with high probability degenerate minimizers which are typically global minimizers.","PeriodicalId":55299,"journal":{"name":"Bulletin of the Polish Academy of Sciences-Technical Sciences","volume":null,"pages":null},"PeriodicalIF":1.2,"publicationDate":"2023-11-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"81102459","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 8

Manufacturing of ZrO2-Ni graded composites via centrifugal casting in the magnetic field 磁场离心铸造制备ZrO2-Ni梯度复合材料

IF 1.2 4区工程技术

Bulletin of the Polish Academy of Sciences-Technical Sciences Pub Date : 2023-11-06 DOI: 10.24425/BPASTS.2020.133379

M. Wachowski, R. Kosturek, H. Winkler, A. Miazga, P. Lada, W. Kaszuwara, K. Konopka, J. Zygmuntowicz

引用次数: 2

On speeding up nano- and micromechanical calculations for irregular systems with long-range potentials 加快具有长程势的不规则体系的纳微力学计算

IF 1.2 4区工程技术

Bulletin of the Polish Academy of Sciences-Technical Sciences Pub Date : 2023-11-06 DOI: 10.24425/BPASTS.2020.131847

L. Rybarska-Rusinek, E. Rejwer, A. Linkov

引用次数: 1

The impact of the COVID-19 pandemic on cybercrime COVID-19大流行对网络犯罪的影响

IF 1.2 4区工程技术

Bulletin of the Polish Academy of Sciences-Technical Sciences Pub Date : 2023-11-06 DOI: 10.24425/bpasts.2021.137933

A. Gryszczyńska

引用次数: 8

Current-fed quasi-Z-source H7 inverter with reduced stress on SiC power devices 减小SiC功率器件应力的准z源电流型H7逆变器

IF 1.2 4区工程技术

Bulletin of the Polish Academy of Sciences-Technical Sciences Pub Date : 2023-11-06 DOI: 10.24425/BPASTS.2019.131568

Przemysław Trochimiuk, M. Zdanowski, J. Rąbkowski

{"title":"Current-fed quasi-Z-source H7 inverter with reduced stress on SiC power devices","authors":"Przemysław Trochimiuk, M. Zdanowski, J. Rąbkowski","doi":"10.24425/BPASTS.2019.131568","DOIUrl":"https://doi.org/10.24425/BPASTS.2019.131568","url":null,"abstract":". This paper discusses selected problems regarding a high-frequency improved current-fed quasi-Z-source inverter (iCFqZSI) designed and built with SiC power devices. At first, new, modified topology of the impedance network is presented. As the structure is derived from the series connection of two networks, the voltage stress across the SiC diodes and the inductors is reduced by a factor of two. Therefore, the SiC MOSFETs may be switched with frequencies above 100 kHz and volume and weight of the passive components is decreased. Furthermore, additional leg with two SiC MOSFETs working as a bidirectional switch is added to limit the current stress during the short-through states. In order to verify the performance of the proposed solution a 6 kVA laboratory model was designed to connect a 400 V DC source (battery) and a 3 £ 400 V grid. According to presented simulations and experimental results high-frequency iCFqZSI is bidirectional – it may act as an inverter, but also as a rectifier. Performed measurements show correct operation at switching frequency of 100 kHz, high quality of the input and output waveforms is observed. The additional leg increases efficiency by up to 0.6% – peak value is 97.8%.","PeriodicalId":55299,"journal":{"name":"Bulletin of the Polish Academy of Sciences-Technical Sciences","volume":null,"pages":null},"PeriodicalIF":1.2,"publicationDate":"2023-11-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"77957114","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Global stability of nonlinear feedback systemswith positive descriptor linear parts 具有正广义线性部分的非线性反馈系统的全局稳定性

IF 1.2 4区工程技术

Bulletin of the Polish Academy of Sciences-Technical Sciences Pub Date : 2023-11-06 DOI: 10.24425/BPAS.2019.127337

T. Kaczorek

引用次数: 3

Computational gait analysis for post-stroke rehabilitation purposes using fuzzy numbers, fractal dimension and neural networks 基于模糊数、分形维数和神经网络的脑卒中后康复步态分析

IF 1.2 4区工程技术

Bulletin of the Polish Academy of Sciences-Technical Sciences Pub Date : 2023-11-06 DOI: 10.24425/BPASTS.2020.131843

Piotr Prokopowicz, D. Mikołajewski, K. Tyburek, E. Mikołajewska

{"title":"Computational gait analysis for post-stroke rehabilitation purposes using fuzzy numbers, fractal dimension and neural networks","authors":"Piotr Prokopowicz, D. Mikołajewski, K. Tyburek, E. Mikołajewska","doi":"10.24425/BPASTS.2020.131843","DOIUrl":"https://doi.org/10.24425/BPASTS.2020.131843","url":null,"abstract":"Computational gait analysis constitutes a useful tool for quantitative assessment of gait disturbances, improving functional diag nosis, assessment of treatment planning, and monitoring of disease progress. There is little research on use of computational gait analysis in neurorehabilitation of post-stroke survivors, but current evidence on its clinical application supports a favorable cost-benefit ratio. The research was conducted among 50 adult people: 25 of them after ischemic stroke constituted the study group, and 25 healthy volunteers constituted the reference group. Study group members were treated for 2 weeks (10 neurorehabilitation sessions). Spatio-temporal gait parameters were assessed before and after therapy and compared using a novel fuzzy-based assessment tool, fractal dimension measurement and gait classification based on artificial neural networks. Measured results of rehabilitation (changes of gait parameters) were statistically relevant and reflected recovery. There is good evidence to extend its use to patients with various gait diseases undergoing neurorehabilitation. However, methodology for properly conducting and interpreting the proposed assessment and analysis procedures, providing validity and reliability of their results remains a key issue. More objective clinical reasoning, based on proposed novel tools, requires further research.","PeriodicalId":55299,"journal":{"name":"Bulletin of the Polish Academy of Sciences-Technical Sciences","volume":null,"pages":null},"PeriodicalIF":1.2,"publicationDate":"2023-11-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"80444696","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 17

LQG/LTR control of input-delayed discrete-time systems 输入延迟离散系统的LQG/LTR控制

IF 1.2 4区工程技术

Bulletin of the Polish Academy of Sciences-Technical Sciences Pub Date : 2023-11-06 DOI: 10.24425/BPASTS.2019.130895

D. Horla, A. Królikowski

引用次数: 3

Scheduling trucks in a multi-door cross-docking system with time windows 带时间窗的多门交叉对接系统中的货车调度

IF 1.2 4区工程技术

Bulletin of the Polish Academy of Sciences-Technical Sciences Pub Date : 2023-11-06 DOI: 10.24425/BPAS.2019.128113

G. Ozden, Inci Sariçiçek

引用次数: 2

Stationary supercapacitor energy storage operation algorithm based on neural network learning system 基于神经网络学习系统的固定式超级电容器储能运行算法

IF 1.2 4区工程技术

Bulletin of the Polish Academy of Sciences-Technical Sciences Pub Date : 2023-11-06 DOI: 10.24425/BPASTS.2020.134176

W. Jefimowski, A. Nikitenko, Z. Drążek, M. Wieczorek

{"title":"Stationary supercapacitor energy storage operation algorithm based on neural network learning system","authors":"W. Jefimowski, A. Nikitenko, Z. Drążek, M. Wieczorek","doi":"10.24425/BPASTS.2020.134176","DOIUrl":"https://doi.org/10.24425/BPASTS.2020.134176","url":null,"abstract":"The paper proposes to apply an algorithm for predicting the minimum level of the state of charge (SoC) of stationary supercapacitor energy storage system operating in a DC traction substation, and for changing it over time. This is done to insure maximum energy recovery for trains while braking. The model of a supercapacitor energy storage system, its algorithms of operation and prediction of the minimum state of charge are described in detail; the main formulae, graphs and results of simulation are also provided. It is proposed to divide the SoC curve into equal periods of time during which the minimum states of charge remain constant. To predict the SoC level for the subsequent period, the learning algorithm based on the neural network could be used. Then, the minimum SoC could be based on two basic types of data: the first one is the time profile of the energy storage load during the previous period with the constant minimum SoC retained, while the second one relies on the trains’ locations and speed values in the previous period. It is proved that the use of variable minimum SoC ensures an increase of the energy volume recovered by approximately 10%. Optimum architecture and activation function of the neural network are also found.","PeriodicalId":55299,"journal":{"name":"Bulletin of the Polish Academy of Sciences-Technical Sciences","volume":null,"pages":null},"PeriodicalIF":1.2,"publicationDate":"2023-11-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"76072366","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 5