Article
Computer Science, Artificial Intelligence
Gao Huang, Zhuang Liu, Geoff Pleiss, Laurens van der Maaten, Kilian Q. Weinberger
Summary: Recent work has shown that adding shorter connections in convolutional networks can make the network deeper, more accurate, and more efficient in training. This paper introduces Dense Convolutional Network (DenseNet), which connects each layer to every other layer in a feed-forward manner. DenseNets alleviate the vanishing-gradient problem, encourage feature reuse, and improve parameter efficiency, leading to significant improvements in object recognition tasks.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2022)
Article
Environmental Sciences
Janne Mayra, Sarita Keski-Saari, Sonja Kivinen, Topi Tanhuanpaa, Pekka Hurskainen, Peter Kullberg, Laura Poikolainen, Arto Viinikka, Sakari Tuominen, Timo Kumpula, Petteri Vihervaara
Summary: Over the past two decades, forest monitoring and inventory systems have shifted from field surveys to remote sensing-based methods. Current analysis methods have limitations, with deep learning methods showing greater potential. This study focused on the classification of major tree species and achieved high accuracy using 3D convolutional neural networks in hyperspectral data.
REMOTE SENSING OF ENVIRONMENT
(2021)
Article
Computer Science, Artificial Intelligence
Saksham Jain, Gautam Seth, Arpit Paruthi, Umang Soni, Girish Kumar
Summary: The study introduces a method of data augmentation using Generative Adversarial Networks, which significantly improves the performance of Convolutional Neural Networks in surface defect classification tasks. Training with synthetic images leads to better classification results.
JOURNAL OF INTELLIGENT MANUFACTURING
(2022)
Article
Multidisciplinary Sciences
Sam Gelman, Sarah A. Fahlberg, Pete Heinzelman, Philip A. Romero, Anthony Gitter
Summary: This study presents a supervised deep learning framework for mapping protein sequence to function, demonstrating superior performance in predicting the behavior of protein sequence variants. Analysis of the trained models highlights the importance of capturing nonlinear interactions and parameter sharing in neural networks for learning the relationship between sequence and function. Additionally, the research shows the networks' ability to learn biologically meaningful information about protein structure and mechanism, as well as design new proteins beyond the training set.
PROCEEDINGS OF THE NATIONAL ACADEMY OF SCIENCES OF THE UNITED STATES OF AMERICA
(2021)
Article
Radiology, Nuclear Medicine & Medical Imaging
Nasim Sirjani, Mostafa Ghelich Oghli, Mohammad Kazem Tarzamni, Masoumeh Gity, Ali Shabanzadeh, Payam Ghaderi, Isaac Shiri, Ardavan Akhavan, Mehri Faraji, Mostafa Taghipour
Summary: Breast cancer is a major cause of cancer-related deaths in women, and early diagnosis is crucial. This study developed a novel deep-learning architecture based on the InceptionV3 network to classify ultrasound breast lesions. The improved InceptionV3 showed robust classification of breast tumors, potentially reducing the need for biopsy in many cases.
PHYSICA MEDICA-EUROPEAN JOURNAL OF MEDICAL PHYSICS
(2023)
Article
Computer Science, Artificial Intelligence
Ronghua Shang, Jiaming Wang, Licheng Jiao, Xiaohui Yang, Yangyang Li
Summary: This paper proposes a spatial feature-based convolutional neural network (SF-CNN) for solving PolSAR classification problems. The special structure of SF-CNN can expand the training set by combining different samples and enhance the network's ability to extract discriminative features in low-dimensional feature space. Experimental results show that SF-CNN outperforms standard CNN in PolSAR image classification tasks.
APPLIED SOFT COMPUTING
(2022)
Article
Engineering, Electrical & Electronic
Kang Zheng, Haiying Wang, Feng Qin, Changhong Miao, Zhigang Han
Summary: This article proposes an improved model based on DeepLab V3+ network and GauGAN data enhancement strategy to address the imbalance in land use datasets and enhance accuracy. By optimizing the generator and discriminator of GauGAN, as well as modifying the modules and feature fusion of DeepLab V3+ network, high-precision land use classification is achieved.
IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING
(2023)
Article
Chemistry, Analytical
Jonghong Kim, Wonhee Lee, Sungdae Baek, Jeong-Ho Hong, Minho Lee
Summary: This paper proposes an incremental learning framework to address the catastrophic forgetting problem in deep neural networks. The framework incorporates the hippocampal memory process and incremental QR factorization to represent feature distribution and reduce forgetting. The experimental results demonstrate that the proposed method effectively improves the stability and plasticity dilemma in deep neural networks.
Article
Chemistry, Analytical
Jaehyeop Choi, Chaehyeon Lee, Donggyu Lee, Heechul Jung
Summary: Modern data augmentation methods like Mixup, CutMix, and the proposed SalfMix have shown improved performance in image recognition tasks, especially when combined as in HybridMix. These methods outperformed traditional single image-based approaches and achieved state-of-the-art results in various classification and object detection datasets.
Article
Biotechnology & Applied Microbiology
Yonglin Zhang, Qi Mo, Li Xue, Jiesi Luo
Summary: Transcription factors (TFs) are key players in gene regulation, and the hybrid CNN + DNN model has shown the best performance in predicting DNA-TF binding specificity. Further research in this area could have broader applications in modeling and predicting TF binding specificity.
Article
Chemistry, Multidisciplinary
Jinho Park, Heegwang Kim, Joonki Paik
Summary: The proposed CF-CNN utilizes a disjoint grouping method to learn multilabel classes, improving classification accuracy by up to 3% with a smaller number of parameters.
APPLIED SCIENCES-BASEL
(2021)
Article
Chemistry, Multidisciplinary
Keisuke Manabe, Yusuke Asami, Tomonari Yamada, Hiroyuki Sugimori
Summary: This study evaluated an improved specialized convolutional neural network for medical image classification tasks, achieving high accuracy especially in pancreatic classification. By optimizing the filter size of the convolution layer and max-pooling, accurate results can be quickly obtained.
APPLIED SCIENCES-BASEL
(2021)
Article
Biotechnology & Applied Microbiology
Pauline Shan Qing Yeoh, Khin Wee Lai, Siew Li Goh, Khairunnisa Hasikin, Xiang Wu, Pei Li
Summary: This study investigates the feasibility of using well-known convolutional neural network (CNN) structures (ResNet, DenseNet, VGG, and AlexNet) to distinguish knees with and without osteoarthritis (OA). The results show that 3D convolutional neural networks using 3D convolutional layers have potential in knee osteoarthritis diagnosis. Transfer learning by transforming 2D pre-trained weights into 3D enhances the performance of the models. This study suggests the possibility of clinical diagnostic aid for knee osteoarthritis using 3DCNN.
FRONTIERS IN BIOENGINEERING AND BIOTECHNOLOGY
(2023)
Article
Environmental Sciences
Jiangbo Xi, Okan K. Ersoy, Ming Cong, Chaoying Zhao, Wei Qu, Tianjun Wu
Summary: This paper proposes a wide and deep Fourier network for efficient feature learning in hyperspectral remote sensing image (HSI) classification. The method utilizes pruned features extracted in the frequency domain to extract hierarchical features layer-by-layer. Experimental results show that the proposed method achieves excellent performance in HSI classification, with the ability to be implemented in lightweight embedded computing platforms.
Review
Environmental Sciences
Abhasha Joshi, Biswajeet Pradhan, Shilpa Gite, Subrata Chakraborty
Summary: Reliable and timely crop-yield prediction and mapping are crucial for food security and decision making. Remote sensing data and deep learning algorithms have been effective tools for crop mapping and yield prediction. This study provides a thorough systematic review of the important scientific works related to state-of-the-art deep learning techniques and remote sensing in crop mapping and yield estimation.
Article
Environmental Sciences
Manuel Carranza-Garcia, Jesus Torres-Mateo, Pedro Lara-Benitez, Jorge Garcia-Gutierrez
Summary: In this study, the performance of existing 2D detection systems for self-driving vehicles on a multi-class problem was evaluated and compared in different scenarios. Despite the increasing popularity of one-stage detectors, it was found that two-stage detectors still provide the most robust performance.
Article
Chemistry, Multidisciplinary
Haotian Wen, Jose Maria Luna-Romera, Jose C. Riquelme, Christian Dwyer, Shery L. Y. Chang
Summary: The morphology of nanoparticles plays a crucial role in determining their properties for various applications. Transmission electron microscopy (TEM) is an effective technique for characterizing nanoparticle morphology at atomic resolution. Developing efficient and automated methods for statistically significant particle metrology is essential for advancing precise particle synthesis and property control.
Article
Computer Science, Artificial Intelligence
Manuel Carranza-Garcia, F. Javier Galan-Sales, Jose Maria Luna-Romera, Jose C. Riquelme
Summary: This paper proposes a novel data fusion architecture for object detection in autonomous driving, using camera and LiDAR data to achieve reliable performance. With deep learning models and sensor data, our approach significantly outperforms previous methods in various weather and lighting conditions.
INTEGRATED COMPUTER-AIDED ENGINEERING
(2022)
Article
Mathematics, Applied
Pedro Lara-Benitez, Manuel Carranza-Garcia, David Gutierrez-Aviles, Jose C. Riquelme
Summary: This study aims to evaluate the performance of different types of deep learning architectures for data streaming classification. The results indicate that convolutional architectures achieve higher accuracy and efficiency but are also most sensitive to concept drifts.
LOGIC JOURNAL OF THE IGPL
(2023)
Article
Chemistry, Multidisciplinary
Laura Madrid-Marquez, Cristina Rubio-Escudero, Beatriz Pontes, Antonio Gonzalez-Perez, Jose C. Riquelme, Maria E. Saez
Summary: This study introduces a new software tool called MOMIC, which provides a complete analysis environment for analyzing and integrating multi-omics data on a single, easy-to-use platform. It offers high editability, reproducibility, and is of great importance for deriving meaningful biological knowledge.
APPLIED SCIENCES-BASEL
(2022)
Article
Energy & Fuels
Tomas Cabello-Lopez, Manuel Carranza-Garcia, Jose C. Riquelme, Jorge Garcia-Gutierrez
Summary: Renewable energies, such as solar power, offer a clean and cost-effective energy source, but their integration into national electricity grids poses challenges due to their dependence on climate and geography. Accurate national-level forecasting is crucial for optimizing energy management, informing policy development, and promoting environmental sustainability. This study aims to address these challenges by improving the accuracy of existing forecasting approaches.
Article
Computer Science, Artificial Intelligence
Pedro Lara-Benitez, Manuel Carranza-Garcia, Jose Maria Luna-Romera, Jose C. Riquelme
Summary: Solar energy is a common and promising source of renewable energy. This work proposes a novel data streaming method for real-time solar irradiance forecasting, using deep learning models. The experiments demonstrate the suitability of deep learning models, particularly MLP and CNN, for this problem.
Proceedings Paper
Computer Science, Artificial Intelligence
Tomas Cabello-Lopez, Manuel Canizares-Juan, Manuel Carranza-Garcia, Jorge Garcia-Gutierrez, Jose C. Riquelme
Summary: This study analyzed wind energy generation data from the Spanish power grid and evaluated the improvement in forecasting quality by detecting concept drifts and retraining models. The experimental results showed that the concept drift approach significantly improved the accuracy of forecasting.
HYBRID ARTIFICIAL INTELLIGENT SYSTEMS, HAIS 2022
(2022)
Proceedings Paper
Computer Science, Artificial Intelligence
Manuel Carranza-Garcia, Pedro Lara-Benitez, Jose Maria Luna-Romera, Jose C. Riquelme
Summary: The importance of feature selection for forecasting solar irradiance time series using spatio-temporal data is studied, and it is found that proper feature selection significantly enhances the forecasting accuracy.
16TH INTERNATIONAL CONFERENCE ON SOFT COMPUTING MODELS IN INDUSTRIAL AND ENVIRONMENTAL APPLICATIONS (SOCO 2021)
(2022)
Proceedings Paper
Computer Science, Artificial Intelligence
Jose Maria Luna-Romera, Manuel Carranza-Garcia, David Gutierrez-Aviles, Jose C. Riquelme-Santos
Summary: This article analyses 5,567 households in London using clustering techniques to characterize different consumer profiles based on their electricity consumption patterns. It suggests adjusting tariffs to reduce consumption peaks and provides insights into the diversity of household consumption.
16TH INTERNATIONAL CONFERENCE ON SOFT COMPUTING MODELS IN INDUSTRIAL AND ENVIRONMENTAL APPLICATIONS (SOCO 2021)
(2022)