Product Design, Manufacturing & Innovation Resources

최신 논문 – 신경망 관련 특허

Neural Networks

팁: 아래 선택 항목 외에도, 저희가 제공하는 두 개의 전체 데이터베이스를 검색하고 필터링할 수 있습니다.

> 무료 간행물 검색 도구 < 저자, 주제, 키워드, 날짜 또는 저널별로 검색.

> 무료 특허 검색 도구 < 유럽 ​​특허청의 영문 특허를 참조하십시오.

본 자료는 전 세계 과학 온라인 저널에 게재된 신경망 관련 영문 논문 및 특허를 엄선하여 정리한 최신 자료입니다. 신경망, 인공 뉴런, 에포크, 신경망 아키텍처, 머신 러닝, 딥 러닝, 서포트 벡터 머신 등의 키워드를 중심으로 분류했습니다.

A residual graph attention network for smart grid stability prediction with incomplete data

Published on 2026-10-20 by @OXFORD

Abstract: AbstractAccurate and reliable stability prediction is an important direction in smart grid, especially when data is incomplete. Traditional machine learning methods usually need complete feature inputs and rely on interpolation techniques, which will distort the relationships between variables. To address this problem, we proposed a graph-based framework that can directly receive incomplete samples without the need for data interpolation. In the proposed method, each sample is regarded as a grap[...]


Our summary: The proposed method utilizes a residual graph attention network to predict smart grid stability from incomplete data. It treats each sample as a graph, leveraging the Pearson correlation coefficient for edge weights. Experimental results demonstrate superior performance compared to traditional models, achieving a recall rate of 0.9340.

residual graph attention network, smart grid stability, incomplete data, feature interactions

Publication

A method to measure properties of a forest

Patent published on the 2026-10-18 in FI under Ref FI20255344 by OY ARBONAUT LTD [FI] (LeppÄnen Vesa [fi], Jeskanen Jeremias [fi], Sener Ecre [fi])

Abstract: A method of determining properties of forest is disclosed. Accurate properties of forest are important for management of forest, forest operations and forest protection. Producing cross-section images of forest allows image recognition techniques to be used in identification of trees (40, 50). Machine learning methods may be used as one form of image recognition techniques. Tree attributes, like size, height or species may be identified for the identified trees. Having overlapping area in the cr[...]


Our summary: A method for measuring forest properties is disclosed. Cross-section images facilitate tree identification using image recognition techniques. Machine learning aids in recognizing tree attributes and managing duplicate identifications.

forest measurement, image recognition, machine learning, tree attributes

Patent

Insulin resistance and brain sulcal morphology in non-diabetic adults

Published on 2026-10-16 by @MIT

Abstract: AbstractInsulin resistance exerts direct effects on brain structure, yet most neuroimaging evidence comes from populations with established diabetes, and conventional gray matter volumetry may lack sensitivity to detect early structural changes. Here, we examined associations between insulin resistance and sulcal morphology, a measure that integrates cortical thinning, subcortical atrophy, and white matter degeneration into a single outcome, in 728 non-diabetic adults (518 men, 210 women; mean a[...]


Our summary: This study investigates the relationship between insulin resistance and sulcal morphology in non-diabetic adults. It finds that higher insulin resistance correlates with increased frontal sulcal volumes, particularly in men under 70. The findings suggest sulcal morphology can reveal early metabolic effects on brain structure not detectable by traditional volumetry.

Insulin resistance, brain morphology, sulcal volume, neuroimaging

Publication

cybersecurity and digitalization in SMEs through the lens of protection motivation theory

Published on 2026-10-16 by @OXFORD

Abstract: AbstractThis study examines the relationship between cybersecurity and the digitalization of small and medium-sized enterprises (SMEs). Digitalization is conceptualized as a gradual, path-dependent process in which firms adopt digital technologies over time, increasing both connectivity and exposure to cyber threats. Drawing on protection motivation theory (PMT), we interpret digitalization as a form of motivated protective behavior shaped by two cognitive mechanisms: threat appraisal and coping[...]


Our summary: This study explores how cybersecurity influences the digitalization of SMEs using protection motivation theory. It finds that cybersecurity practices positively correlate with digitalization, influenced by SMEs digital maturity. Early digitalization is driven by perceived cyber risk, while established cybersecurity routines facilitate further digital adoption.

cybersecurity, digitalization, protection motivation theory, SMEs

Publication

Development of the social perception pathway across the third trimester of gestation and perinatal transition

Published on 2026-10-15 by @MIT

Abstract: AbstractBabies display social perceptual skills within hours of birth which may indicate that the neural predecessors of such skills have prenatal origins. However, the putative fetal neural infrastructure supporting these skills remains unexplored. Data from 497 participants scanned between 28–44 weeks postmenstrual age were examined to map the development of intrinsic functional connectivity (iFC) within the proto social pathway (pSP), encompassing the superior temporal sulcus (STS), a regio[...]


Our summary: The study maps the development of intrinsic functional connectivity within the proto social pathway and proto ventral pathway during the third trimester. Significant increases in connectivity were observed, indicating early neural architecture for social perception. Findings suggest that key features of the social brain begin to develop prenatally, laying the groundwork for later social skills.

social perception, intrinsic functional connectivity, prenatal development, neural architecture

Publication

Generating query language using a graph based interlingual representation

Patent published on the 2026-10-15 in US under Ref US20260320038 by INT BUSINESS MACHINES CORPORATION [US] (Shu Xiaokui [us], Eykholt Kevin [us], Park Youngja [us], Jang Jiyong [us])

Abstract: A computer-implemented method for transforming natural language to a target query language for performing a cybersecurity task is provided. A processor set receives a request comprises natural language description of the cybersecurity task. The processor set converts the natural language description of the cybersecurity task into a graph using a machine learning model. The processor set translates the graph to a cybersecurity query in the target query language using a compiler. The processor set[...]


Our summary: A method converts natural language descriptions of cybersecurity tasks into a graph representation. This graph is then translated into a target query language using a compiler. Finally, the generated query is executed on a cybersecurity platform to perform the task.

interlingual representation, natural language processing, machine learning, cybersecurity query

Patent

Mitigating reward hacking by training generative neural network systems

Patent published on the 2026-10-15 in WO under Ref WO2026217788 by GDM HOLDING LLC [US] (Kassraie Parnian [ch], Sessa Pier Giuseppe [ch], Bachem Olivier FrÉdÉric [ch], Hussenot Desenonges LÉonard [fr], RamÉ Alexandre Hippolyte Candide Marie [fr], Ferret Johan [fr], Girgin Sertan [fr], Sinopalnikov Danila [ch], Momchev Nikola [ch])

Abstract: Methods, systems, and computer program code for training, e.g. fine-tuning, a generative neural network system. The method involves processing a training input sequence using generative and reference neural network systems to generate a respective first and second output sequences. These are evaluated by a test reward model to obtain respective first and second rewards that are used to identify the presence of a risk of reward hacking. If a risk of reward hacking is present the training is adjus[...]


Our summary: This content discusses methods for training generative neural networks while mitigating reward hacking. It involves processing input sequences through neural networks and evaluating outputs with a test reward model. Adjustments to training are made when risks of reward hacking are detected, enhancing output quality.

reward hacking, generative neural networks, training methods, risk mitigation

Patent

Information processing program, information processing method, information processing system, and information processing terminal

Patent published on the 2026-10-15 in WO under Ref WO2026216405 by MINEBEA MITSUMI INC [JP] (Fuse Tohru [jp], Sumioka Jun [jp])

Abstract: This information processing program causes a computer to: cause a vibration generation device (vibration generation unit 22) provided in a terminal (user terminal 1) to generate vibration; cause a sensor (vibration measurement unit 24) provided in the terminal to measure the vibration; cause a model (machine learning model 40), which is trained by machine learning using training data including vibration data to output a state including at least one of a contact state and a storage state of the t[...]


Our summary: The program enables a terminal to generate vibrations and measure them. It uses a machine learning model to analyze the vibration data. The system provides notifications based on the analyzed state of the terminal.

information processing, vibration generation, machine learning, sensor measurement

Patent

다룬 주제: 신경망, 인공 뉴런, 에포크, 신경망 아키텍처, 머신 러닝, 딥 러닝, 서포트 벡터 머신, 동적 손동작 인식, 벤치마크 데이터셋, 사전 학습된 가중치, 훈련 데이터, 훈련 데이터셋, 설계 스냅샷, 수상돌기 뉴런 모델, 시냅스 가소성, 강화 동적 그룹화 미분 진화, YOLOv8, ByteTrack.

역사적 맥락

2000
2002
2010
2013
2000
2000
2003
2010
2013-09-24

(날짜를 알 수 없거나 관련이 없는 경우, 예를 들어 "유체역학"의 경우, 주목할 만한 등장 시기를 대략적으로 추정하여 제공합니다.)

10년차 기념일

고화질 이미지 및 다운로드는 등록된 회원에게만 100% 무료로 제공됩니다.

> 로그인 <