Deep Contrastive Representation Learning With Self-Distillation

IEEE Transactions on Emerging Topics in Computational Intelligence, Год журнала: 2023, Номер 8(1), С. 3 - 15

Опубликована: Авг. 29, 2023

Recently, contrastive learning (CL) is a promising way of discriminative representations from time series data. In the representation hierarchy, semantic information extracted lower levels basis that captured higher levels. Low-level essential and should be considered in CL process. However, existing algorithms mainly focus on similarity high-level information. Considering low-level may improve performance CL. To this end, we present deep with self-distillation (DCRLS) for domain. DCRLS gracefully combine data augmentation, learning, self distillation. Our augmentation provides different views same sample as input DCRLS. Unlike most concentrate only, our also considers contrast between peer residual blocks. distillation promotes knowledge flow to blocks help regularize transfer The experimental results demonstrate DCRLS-based structures achieve excellent classification clustering 36 UCR2018 datasets.

Язык: Английский

---

DTCM: Deep Transformer Capsule Mutual Distillation for Multivariate Time Series Classification

IEEE Transactions on Cognitive and Developmental Systems, Год журнала: 2024, Номер 16(4), С. 1445 - 1461

Опубликована: Фев. 26, 2024

This paper proposes a dual-network-based feature extractor, perceptive capsule network (PCapN), for multivariate time series classification (MTSC), including local (LFN) and global relation (GRN). The LFN has two heads (i.e., Head_A Head_B), each containing squash CNN blocks one dynamic routing block to extract the features from data mine connections among them. GRN consists of capsule-based transformer capture patterns variable correlate useful information multiple variables. Unfortunately, it is difficult directly deploy PCapN on mobile devices due its strict requirement computing resources. So, this designs lightweight (LCapN) mimic cumbersome PCapN. To promote knowledge transfer LCapN, deep mutual (DTCM) distillation method. It targeted offline, using one- two-way operations supervise process student teacher models. Experimental results show that proposed DTCM achieve excellent performance UEA2018 datasets regarding top-1 accuracy.

Язык: Английский

---

Enhancing heart disease prediction using a self-attention-based transformer model

Scientific Reports, Год журнала: 2024, Номер 14(1)

Опубликована: Янв. 4, 2024

Abstract Cardiovascular diseases (CVDs) continue to be the leading cause of more than 17 million mortalities worldwide. The early detection heart failure with high accuracy is crucial for clinical trials and therapy. Patients will categorized into various types disease based on characteristics like blood pressure, cholesterol levels, rate, other characteristics. With use an automatic system, we can provide diagnoses those who are prone by analyzing their In this work, deploy a novel self-attention-based transformer model, that combines self-attention mechanisms networks predict CVD risk. layers capture contextual information generate representations effectively model complex patterns in data. Self-attention interpretability giving each component input sequence certain amount attention weight. This includes adjusting output layers, incorporating modifying processes collect relevant information. also makes it possible physicians comprehend which features data contributed model's predictions. proposed tested Cleveland dataset, benchmark dataset University California Irvine (UCI) machine learning (ML) repository. Comparing several baseline approaches, achieved highest 96.51%. Furthermore, outcomes our experiments demonstrate prediction rate higher cutting-edge approaches used prediction.

Язык: Английский

---

Fault diagnosis study of hydraulic pump based on improved symplectic geometry reconstruction data enhancement method

Advanced Engineering Informatics, Год журнала: 2024, Номер 61, С. 102459 - 102459

Опубликована: Март 5, 2024

Язык: Английский

---

Segmentation and Phenotype Calculation of Rapeseed Pods Based on YOLO v8 and Mask R-Convolution Neural Networks

Plants, Год журнала: 2023, Номер 12(18), С. 3328 - 3328

Опубликована: Сен. 20, 2023

Rapeseed is a significant oil crop, and the size length of its pods affect productivity. However, manually counting number rapeseed measuring length, width, area pod takes time effort, especially when there are hundreds resources to be assessed. This work created two state-of-the-art deep learning-based methods identify related attributes, which then implemented in pots improve accuracy yield estimate. One these YOLO v8, other two-stage model Mask R-CNN based on framework Detectron2. The v8n with Resnet101 backbone Detectron2 both achieve precision rates exceeding 90%. recognition results demonstrated that models perform well graphic images segmented. In light this, we developed coin-based approach for estimating tested it test dataset made up nine different species Brassica napus one campestris L. correlation coefficients between manual measurement machine vision width were calculated using statistical methods. regression coefficient was 0.991, 0.989. conclusion, first time, utilized learning techniques characteristics while concurrently establishing pods. Our suggested approaches successful segmenting precisely. offers breeders an effective strategy digitally analyzing phenotypes automating identification screening process, not only germplasm but also leguminous plants, like soybeans possess

Язык: Английский

---

STF-YOLO: A small target detection algorithm for UAV remote sensing images based on improved SwinTransformer and class weighted classification decoupling head

Measurement, Год журнала: 2023, Номер 224, С. 113936 - 113936

Опубликована: Ноя. 26, 2023

Язык: Английский

---

A Systematic Review on Deep Learning with CNNs Applied to Surface Defect Detection

Journal of Imaging, Год журнала: 2023, Номер 9(10), С. 193 - 193

Опубликована: Сен. 25, 2023

Surface defect detection with machine learning has become an important tool in industries and a large field of study for researchers or workers recent years. It is necessary to have simplified source information that helps us better focus on one type surface. In this systematic review, we present classification surface based convolutional neural networks (CNNs) focused types. Findings: Out 253 records identified, 59 primary studies were eligible. Following the Preferred Reporting Items Systematic Reviews Meta-Analyses (PRISMA) guidelines, analyzed structures each concepts related defects their types surfaces. The presented review mainly finding surfaces most used industry (metal, building, ceramic, wood, special). We delve into specifics category, offering illustrative examples applications within both industrial laboratory settings. Furthermore, propose new taxonomy obtained results collected information. summarized extracted main characteristics such as surface, problem types, timeline, network, techniques, datasets. Among relevant our analysis, found metallic used, it 62.71% studies, prevalent classification, accounting 49.15% total. observe transfer was employed 83.05% while data augmentation utilized 59.32%. Our findings also provide insights cameras frequently employed, along strategies adopted address illumination challenges certain articles approach creating datasets real-world applications. allow quick efficient search professionals interested improving projects. Finally, trends could open fields future research area detection.

Язык: Английский

---

MED-YOLOv8s: a new real-time road crack, pothole, and patch detection model

Journal of Real-Time Image Processing, Год журнала: 2024, Номер 21(2)

Опубликована: Янв. 29, 2024

Язык: Английский

---

From global challenges to local solutions: A review of cross-country collaborations and winning strategies in road damage detection

Advanced Engineering Informatics, Год журнала: 2024, Номер 60, С. 102388 - 102388

Опубликована: Март 7, 2024

Язык: Английский

---

A regularized constrained two-stream convolution augmented Transformer for aircraft engine remaining useful life prediction

Engineering Applications of Artificial Intelligence, Год журнала: 2024, Номер 133, С. 108161 - 108161

Опубликована: Март 11, 2024

Язык: Английский

---