Article
Multidisciplinary Sciences
Zhuangwen Wu, Zhiping Wan, Dongdong Ge, Ludan Pan
Summary: This study proposes a method for recognizing car engine sounds based on a deformable feature map residual network. It extracts the time-frequency image features using offset and convolutional layers, and fuses them with the Mel frequency cepstral coefficients. Experimental results show that the proposed method achieves a significantly higher accuracy under various operating conditions compared to existing methods.
SCIENTIFIC REPORTS
(2022)
Article
Multidisciplinary Sciences
Adityanarayanan Radhakrishnan, Mikhail Belkin, Caroline Uhler
Summary: This study investigates the consistency problem of neural networks trained using standard procedures for classification tasks and constructs a set of explicit consistent neural network classifiers. The research finds that deep networks have significant advantages in classification tasks, while excessive depth can be harmful in regression tasks.
PROCEEDINGS OF THE NATIONAL ACADEMY OF SCIENCES OF THE UNITED STATES OF AMERICA
(2023)
Article
Computer Science, Artificial Intelligence
Jialin Liu, Fei Chao, Chih-Min Lin, Changle Zhou, Changjing Shang
Summary: This paper introduces dynamic kernel convolutional neural networks (DK-CNNs) and explains how they enhance the expressive capacity of convolutional operations by extending a latent dimension. DK convolution analyzes fixed features with a latent variable, leading to better performance compared to regular CNNs.
Article
Computer Science, Artificial Intelligence
Jakub Zak, Anna Korzynska, Antonina Pater, Lukasz Roszkowiak
Summary: In this study, the combination of the Fourier Transform and Convolutional Neural Network methods was used to classify images in multiple datasets. By incorporating Fourier Transform Layer, the processing speed was increased without sacrificing accuracy, providing an alternative approach to Convolutional Neural Networks that reduces the need for GPU training. Experimental results showed that models with the proposed layer achieved comparable test accuracy to convolutional models for images of size 128 x 128 and larger, with a minimum of 27% reduction in training time per one epoch on Central Processing Unit.
APPLIED SOFT COMPUTING
(2023)
Article
Computer Science, Artificial Intelligence
Lingraj Dora, Sanjay Agrawal, Rutuparna Panda, Ram Bilas Pachori
Summary: This paper proposes a multiple kernel-based convolutional neural network (MK-CNN) approach for automated pathological brain classification task. By using multi-scale features and considering both regional specifics and global spatial consistency, the proposed method outperforms state-of-the-art techniques on real patient data and can aid experts in conducting clinical follow-up studies.
NEURAL COMPUTING & APPLICATIONS
(2023)
Article
Chemistry, Analytical
Olivia Nocentini, Jaeseok Kim, Muhammad Zain Bashir, Filippo Cavallo
Summary: With the increasing elderly population, there is a growing need for caregivers, which may pose challenges for society. To address this, the use of service robotics is proposed as a solution, particularly in household settings and aged people's homes. This paper focuses on the manipulation of clothes, a daily activity, and proposes the study of fashion image classification using neural network models. The results show that the MCNN15 model achieved a classification accuracy of 94.04% on the Fashion-MNIST dataset, outperforming previous literature.
Article
Chemistry, Analytical
Wei Zhang, Weiwei Feng, Zongqi Cai, Huanqing Wang, Qi Yan, Qing Wang
Summary: This paper proposes a method using Raman spectroscopy combined with a one-dimensional convolutional neural network (1D-CNN) to accurately identify microplastics. The experimental results show that this method achieves the highest accuracies of 96.4% and 96.2% based on raw data and pre-processed data, respectively. Compared with K-Nearest Neighbors (KNN), Support Vector Machine (SVM), Multi-layer Perceptron (MLP), and AlexNet classifiers, the proposed method has an average accuracy of 95.8% and 95.5% on raw data and pre-processed data, respectively. This study proves that Raman spectroscopy combined with 1D-CNN can effectively and accurately classify microplastics, regardless of whether the spectra data are pre-processed or not, and can further shorten the recognition time, providing a reference for future microplastics detection.
VIBRATIONAL SPECTROSCOPY
(2023)
Article
Engineering, Biomedical
Zhuyao Fan, Xugang Xi, Yunyuan Gao, Ting Wang, Feng Fang, Michael Houston, Yingchun Zhang, Lihua Li, Zhong Lu
Summary: This study proposes a novel algorithm by inserting two modules into CNN to solve the problem that traditional 1D-CNNs are unable to acquire both frequency domain and channel association information. The proposed algorithm achieved an improvement of 6.6% on the motion imagery 4-classes recognition mission and 11.3% on the 2-classes classification task compared to traditional decoding algorithms.
IEEE TRANSACTIONS ON NEURAL SYSTEMS AND REHABILITATION ENGINEERING
(2023)
Article
Engineering, Electrical & Electronic
Ruimin Song, Weigen Chen, Dingkun Yang, Haiyang Shi, Ruyue Zhang, Zewei Wang
Summary: Detecting insulation performance through testing Raman spectroscopy of transformer oil is noninvasive and shows great promise in on-site diagnosis. The method proposed in this article has the potential to provide accurate determination for diagnosing oil-immersed transformers' aging conditions.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2021)
Article
Computer Science, Artificial Intelligence
El Houssaine Hssayni, Nour-Eddine Joudar, Mohamed Ettaouil
Summary: KRR-CNN, a new optimization model for reducing kernels redundancy in CNN, utilizes an evolutionary genetic algorithm to efficiently reduce redundant kernels and enhance classification performance.
NEURAL COMPUTING & APPLICATIONS
(2022)
Article
Computer Science, Artificial Intelligence
Xinxin Shan, Tai Ma, Yutao Shen, Jiafeng Li, Ying Wen
Summary: The study introduces a novel attention convolution method called Kernel Attention Convolution (KAConv) to enhance the flexibility of convolution. By embedding attention into the convolution kernel, KAConv generates different attention weights to dynamically adjust the parameters of convolution kernels, improving the flexibility of convolution. Experiment results demonstrate that KAConv outperforms existing attention mechanism-based methods in the ImageNet-1K benchmark.
Article
Agriculture, Dairy & Animal Science
Ji Wang, Han Zhang, Nanzhu Chen, Tong Zeng, Xiaohua Ai, Keliang Wu
Summary: This study developed a deep learning framework called PorcineAI-enhancer to predict enhancer sequences in pigs. The model showed excellent performance and strong predictive capability for unknown and tissue-specific enhancers. This research provides valuable resources for future studies on gene expression regulation in pigs.
Article
Computer Science, Artificial Intelligence
Brendan Kolisnik, Isaac Hogan, Farhana Zulkernine
Summary: We propose a hierarchical image classification model, Condition-CNN, which improves prediction accuracy and reduces training time by using the Teacher Forcing training algorithm and conditional probabilities. The validation results show that Condition-CNN achieves higher prediction accuracy for Level 1, 2, and 3 classes compared to other baseline CNN models.
EXPERT SYSTEMS WITH APPLICATIONS
(2021)
Article
Computer Science, Artificial Intelligence
Francesco Ponzio, Enrico Macii, Elisa Ficarra, Santa Di Cataldo
Summary: In real-world scenarios, training Convolutional Neural Networks (CNNs) with high quality images and correct labels is difficult. This affects the performance of CNNs during both training and inference. To tackle this issue, we propose a new two-module CNN called Wise2WipedNet (W2WNet), which uses Bayesian inference to identify and discard spurious images during training and provides prediction confidence during inference. Our experiments on various image classification tasks and histological image analysis demonstrate that W2WNet can effectively identify image degradation and mislabelling issues, resulting in improved classification accuracy.
EXPERT SYSTEMS WITH APPLICATIONS
(2023)
Article
Computer Science, Information Systems
Kwanghyun Koo, Hyun Kim
Summary: In this study, a new vectorized structured kernel pruning method is proposed, which achieves high FLOPs reduction and minimal accuracy degradation while maintaining the weight structure. Experimental results demonstrate significant parameter and FLOPs reduction, as well as real acceleration effects on GPUs, in various networks including ResNet-50.
Article
Chemistry, Analytical
Samjin Choi, Sang Woong Moon, Seung Ho Lee, Wansun Kim, Soogeun Kim, Su Kang Kim, Jae-Ho Shin, Young-Guk Park, Kyung-Hyun Jin, Tae Gi Kim
SENSORS AND ACTUATORS B-CHEMICAL
(2019)
Article
Nanoscience & Nanotechnology
Soogeun Kim, Tae Gi Kim, Soo Hyun Lee, Wansun Kim, Ayoung Bang, Sang Woong Moon, Jeongyoon Song, Jae-Ho Shin, Jae Su Yu, Samjin Choi
ACS APPLIED MATERIALS & INTERFACES
(2020)
Article
Optics
Munsik Choi, Soogeun Kim, Seung Ho Choi, Hyeong-Ho Park, Kyung Min Byun
Article
Spectroscopy
Soogeun Kim, Young Jin Kim, Ayoung Bang, Wansun Kim, Samjin Choi, Hee Joo Lee
Summary: The study investigated the impact of Raman excitation wavelengths on the surface-enhanced Raman spectroscopy (SERS) for the identification of nontuberculous mycobacteria (NTM). By comparing SERS spectra with three commonly used excitation wavelengths, it was found that NTM species could be distinguished with statistical differences. Utilizing principal components-linear discriminant analysis and leave-one-out cross validation, the study achieved identification accuracies for six NTM species using aromatic amino acid biomarkers at different excitation wavelengths.
SPECTROCHIMICA ACTA PART A-MOLECULAR AND BIOMOLECULAR SPECTROSCOPY
(2021)
Article
Chemistry, Analytical
Soogeun Kim, Jin-Ho Joo, Wansun Kim, Ayoung Bang, Hyung Woo Choi, Sang Woong Moon, Samjin Choi
Summary: A high-performance portable Raman scattering-based sensing platform has been developed for the detection of individual toxic chemicals and simultaneous determination of toxic mixtures with significantly low detection limits for methylene blue and malachite green, as well as good performance in determining individual toxic chemicals in complex aquatic mixtures.
SENSORS AND ACTUATORS B-CHEMICAL
(2021)
Article
Chemistry, Physical
Hyun-Jung Kim, Ji-Hyun Jang, Sang Uk Woo, Kyoung-Kyu Choi, Sun-Young Kim, Jack L. Ferracane, Jung-Hwan Lee, Dongseok Choi, Samjin Choi, Soogeun Kim, Ayoung Bang, Duck-Su Kim
Summary: The study found that the novel BAG-containing dentin adhesive has the potential to reduce the permeability of demineralized dentin and promote dentin remineralization.
Article
Optics
Ji Hyeon Choi, Munsik Choi, Tien Son Ho, Soogeun Kim, Samjin Choi, Seung Ho Choi, Kyung Min Byun
Summary: A flexible SERS sensor template with nanogap-rich gold nanoislands was fabricated using a silk fibroin film, which showed good detection performance and investigated the effect of surface roughness of the silk substrate on sensor performance.
Article
Biophysics
Wansun Kim, Soogeun Kim, Jisang Han, Tae Gi Kim, Ayoung Bang, Hyung Woo Choi, Gyeong Eun Min, Jae-Ho Shin, Sang Woong Moon, Samjin Choi
Summary: We introduce a label-free surface-enhanced Raman scattering (SERS) biosensing platform that can identify the efficacy of the Oxford-AstraZeneca (AZD1222) vaccine in vaccinated individuals using non-invasive tear samples. The tears of people who receive the AZD1222 vaccine may be similar to those of adenovirus epidemic keratoconjunctivitis patients, as the vaccine is derived from a replication-deficient ChAdOx1 vector of chimpanzee adenovirus. By analyzing the signals from antibodies or immunoglobulin G by-product, we confirmed the potential of three markers for estimating the vaccination status.
BIOSENSORS & BIOELECTRONICS
(2022)
Article
Chemistry, Physical
Ji-Hyun Jang, Hyun-Jung Kim, Joo-Young Choi, Hae-Won Kim, Samjin Choi, Soogeun Kim, Ayoung Bang, Duck-Su Kim
Summary: The objective of this study was to evaluate the effect of novel bioactive glass (BAG)-containing desensitizers on dentin permeability. The results showed that desensitizers containing BAG significantly reduced dentin fluid flow rates after two weeks of storage in simulated body fluid. Raman spectroscopy analysis also revealed the formation of hydroxyapatite precipitates on the dentin surface after desensitizer treatment.
Article
Biophysics
Wansun Kim, Ayoung Bang, Soogeun Kim, Gi-Ja Lee, Yeon-Hee Kim, Samjin Choi
Summary: The study demonstrates the successful use of an SERS immunoassay platform based on gold nanotriangles for detecting adiponectin levels in the biofluids of pregnant women, showing a wide assay range, high reliability, and excellent selectivity.
BIOSENSORS & BIOELECTRONICS
(2022)
Article
Chemistry, Analytical
Soogeun Kim, Wansun Kim, Ayoung Bang, Jeong-Yoon Song, Jae-Ho Shin, Samjin Choi
Summary: This study introduces a label-free, simple, and high-efficiency breast cancer detection platform with multimodal biomarker analytic algorithms, demonstrating the potential for real-time diagnosis of breast cancer during surgery.
ANALYTICAL METHODS
(2021)
Proceedings Paper
Materials Science, Multidisciplinary
Soogeun Kim, Ayoung Bang, Samjin Choi
NANOENGINEERING: FABRICATION, PROPERTIES, OPTICS, THIN FILMS, AND DEVICES XVI
(2019)
Article
Chemistry, Analytical
Soogeun Kim, Seung Ho Lee, Young Jin Kim, Hee Joo Lee, Samjin Choi
ANALYTICAL METHODS
(2019)