Article
Environmental Sciences
Adrian Doicu, Alexandru Doicu, Dmitry S. Efremenko, Diego Loyola, Thomas Trautmann
Summary: This paper presents neural network methods for predicting uncertainty in atmospheric remote sensing, which include solutions for direct and inverse problems in a Bayesian framework. Methods involve simulating radiative transfer models and using Bayesian approaches, as well as neural networks for predicting uncertainties in noise distributions and input parameters. Testing is done using a neural network with assumed density filtering and interval arithmetic, with analysis focusing on the retrieval of cloud parameters from radiances measured by the Earth Polychromatic Imaging Camera (EPIC).
Article
Computer Science, Theory & Methods
Junkyu Lee, Lev Mukhanov, Amir Sabbagh Molahosseini, Umar Minhas, Yang Hua, Jesus Martinez Del Rincon, Kiril Dichev, Cheol-Ho Hong, Hans Vandierendonck
Summary: This article provides a survey on resource-efficient CNN techniques in terms of model-, arithmetic-, and implementation-level techniques, and discusses the future trend for resource-efficient CNN research.
ACM COMPUTING SURVEYS
(2023)
Article
Chemistry, Analytical
A. Al -Ali, B. Maundy, A. Allagui, A. Elwakil
Summary: A deep learning based algorithm for finding the optimum impedance spectroscopy model is proposed in this study. The algorithm uses convolutional neural networks with long short-term memory to identify the circuit model that best fits the measured spectral impedance data. A modified two-stage optimization technique is also proposed to find the optimal circuit model parameters. The algorithm is validated using experimentally measured battery impedance data and applied to cherry tomato bio-impedance data.
JOURNAL OF ELECTROANALYTICAL CHEMISTRY
(2022)
Article
Computer Science, Artificial Intelligence
Leandro Aparecido S. Passos, Danilo S. Jodas, Luiz C. F. Ribeiro, Marco Akio, Andre Nunes De Souza, Joao Paulo Papa
Summary: In the past decade, machine learning-based approaches have made significant progress and outperformed humans in many complex tasks. This is partially due to the exponential growth in available data, which allows for the extraction of reliable real-world information. However, the imbalanced nature of the data can lead to biased machine learning models. In this paper, three strategies based on the Optimum-Path Forest (OPF) are proposed to address the imbalance problem, and the robustness of these strategies is confirmed through comparisons with state-of-the-art techniques.
KNOWLEDGE-BASED SYSTEMS
(2022)
Article
Agronomy
Yinglun Li, Xiaohai Zhan, Shouyang Liu, Hao Lu, Ruibo Jiang, Wei Guo, Scott Chapman, Yufeng Ge, Benoit de Solan, Yanfeng Ding, Frederic Baret
Summary: In this work, a high-throughput method was developed to count the number of leaves by detecting leaf tips in RGB images. Realism of the images was improved using domain adaptation methods before training deep learning models. The proposed method demonstrated efficiency and the best performance was achieved with the Faster-RCNN model and cycle-consistent generative adversarial network adaptation technique.
Article
Chemistry, Multidisciplinary
Andrzej Piegat, Marcin Plucinski
Summary: This article presents a method for solving control problems with interval coefficients and illustrates its application in optimizing sugar beet fertilization. By introducing the concept of robustness, previously unsolvable problems can be effectively addressed.
APPLIED SCIENCES-BASEL
(2022)
Article
Computer Science, Hardware & Architecture
Vitor Finotti, Bruno Albertini
Summary: The study proposes a method to simulate the effects of quantization in CNN inference, reducing complexity and overhead while enabling fast post-training quantization. Experimental results show significant reductions in model size while maintaining classification accuracy, and explore the relationship between classification complexity and tolerance to quantization.
COMPUTERS & ELECTRICAL ENGINEERING
(2021)
Article
Ecology
Ali Seydi Keceli, Aydin Kaya, Cagatay Catal, Bedir Tekinerdogan
Summary: The manual prediction of plant species and diseases is costly and time-consuming, and expertise may not always be available. Automated approaches, such as machine learning and deep learning, are being used to overcome these challenges. This study proposes a novel multi-task learning approach that combines plant species and disease prediction tasks using shared representations. The results show that this approach improves efficiency and learning speed.
ECOLOGICAL INFORMATICS
(2022)
Article
Computer Science, Interdisciplinary Applications
Bao-Luo Li, Yu Qi, Jian-Sheng Fan, Yu-Fei Liu, Cheng Liu
Summary: Crack identification is crucial for preventive maintenance of asphalt pavement. This paper describes a fusion model based on the YOLO v5 that combines grid-based classification and box-based detection, achieving high accuracy and efficiency. The proposed NMS-ARS algorithm improves crack topology detection through postprocessing. Experimental results demonstrate the effective automatic crack identification for asphalt pavement.
COMPUTER-AIDED CIVIL AND INFRASTRUCTURE ENGINEERING
(2023)
Article
Computer Science, Artificial Intelligence
Mohsen Heidari, Mohammad Hossein Moattar, Hamidreza Ghaffari
Summary: Dropout is a mechanism to prevent overfitting and improve the generalization of deep neural networks. Random dropout randomly terminates nodes in each training step, which may decrease network accuracy. In dynamic dropout, the importance of each node is calculated, and important nodes are not dropped. However, the calculation of node importance is not consistent, and it is costly to calculate the importance in each training step.
Article
Radiology, Nuclear Medicine & Medical Imaging
Gian Marco Conte, Alexander D. Weston, David C. Vogelsang, Kenneth A. Philbrick, Jason C. Cai, Maurizio Barbera, Francesco Sanvito, Daniel H. Lachance, Robert B. Jenkins, W. Oliver Tobin, Jeanette E. Eckel-Passow, Bradley J. Erickson
Summary: Generative adversarial networks can synthesize brain MRI scans to replace missing sequences, enabling their use as inputs for deep learning models in brain lesion segmentation. The quality of the generated images was evaluated based on MSE and SSI, while the segmentation results were compared using DSC, confirming their effectiveness.
Article
Computer Science, Artificial Intelligence
E. Gardini, M. J. Ferrarotti, A. Cavalli, S. Decherchi
Summary: Computational intelligence, particularly deep learning, offers powerful tools for discriminating and generating samples such as images. Research indicates that spaces induced by deep-learning convolutional neural networks can capture historical/stylistic progressions in music and visual art. Experiments conducted with a principal path algorithm in the music and visual art domains show reasonable historical/stylistic progressions when considering a subset of classes.
COGNITIVE COMPUTATION
(2021)
Article
Computer Science, Artificial Intelligence
Junde Chen, Defu Zhang, Adnan Zeb, Yaser A. Nanehkaran
Summary: Rice is a crucial crop, but diseases can harm yields and food security. This study utilizes deep learning and attention mechanisms to optimize model training, achieving efficient rice disease identification with high accuracy.
EXPERT SYSTEMS WITH APPLICATIONS
(2021)
Article
Multidisciplinary Sciences
Pichatorn Suppakitjanusant, Somnuek Sungkanuparph, Thananya Wongsinin, Sirapong Virapongsiri, Nittaya Kasemkosin, Laor Chailurkit, Boonsong Ongphiphadhanakul
Summary: Recent breakthroughs in deep learning have allowed for the detection of subtle changes in voice features of COVID-19 patients post-recovery, with the model using polysyllabic sentences achieving the highest accuracy of 85%.
SCIENTIFIC REPORTS
(2021)
Article
Computer Science, Artificial Intelligence
Philipp Christian Petersen, Anna Sepliarskaia
Summary: This study investigates the generalization capacity of group convolutional neural networks and provides precise estimates of VC dimensions for certain simple sets of such networks. It is discovered that even with infinite groups and suitable convolutional kernels, two-parameter families of convolutional neural networks can have infinite VC dimensions, while remaining invariant to the action of an infinite group.
Article
Computer Science, Theory & Methods
Claudio Filipi Goncalves Dos Santos, Joao Paulo Papa
Summary: This paper analyzes the application of various regularization methods developed in recent years to CNN models, including data augmentation, internal changes, and label transformations, demonstrating significant improvements in different CNN models. What sets this paper apart from other surveys is that it focuses on papers published in the last five years and emphasizes the reproducibility of the results.
ACM COMPUTING SURVEYS
(2022)
Article
Computer Science, Artificial Intelligence
Gustavo H. de Rosa, Mateus Roder, Joao P. Papa
Summary: Biometric recognition has provided direct methods for identifying individuals under specific circumstances and enhancing security policies. This study introduces the use of handwritten dynamics and convolutional neural networks to identify individual identities, resulting in new advancements in the field.
Article
Computer Science, Artificial Intelligence
Luiz C. F. Ribeiro, Gustavo H. de Rosa, Douglas Rodrigues, Joao P. Papa
Summary: This paper proposes creating Convolutional Neural Networks ensembles through Single-Iteration Optimization to address the issue of highly specific hyperparameter settings. The results demonstrate that this method can achieve promising results while reducing the time required.
Article
Computer Science, Information Systems
Zaid Abdi Alkareem Alyasseri, Osama Ahmad Alomari, Sharif Naser Makhadmeh, Seyedali Mirjalili, Mohammed Azmi Al-Betar, Salwani Abdullah, Nabeel Salih Ali, Joao P. Papa, Douglas Rodrigues, Ammar Kamal Abasi
Summary: Electroencephalogram signals (EEG) provide biometric identification systems with unique and universal features. This paper formulates the EEG channel selection problem as a binary optimization problem and uses the BGWO algorithm to find an optimal solution. The proposed method achieves good results in person identification using SVM-RBF classifier and auto-regressive coefficients for feature extraction.
Article
Chemistry, Analytical
Zaid Abdi Alkareem Alyasseri, Osama Ahmad Alomari, Joao P. Papa, Mohammed Azmi Al-Betar, Karrar Hameed Abdulkareem, Mazin Abed Mohammed, Seifedine Kadry, Orawit Thinnukool, Pattaraporn Khuwuthyakorn
Summary: The electroencephalogram (EEG) has great potential for user identification, but selecting which electrodes to use is a challenging task. This study introduces a new algorithm that selects the most representative electrodes using optimization methods, and experimental results show its accuracy.
Article
Plant Sciences
Yutcelia Galviz, Gustavo M. Souza, Ulrich Luettge
Summary: This article introduces the concept that both biotic and abiotic factors can act as stressors. Depending on the strength and duration of stressors, there can be positive eustress and negative distress. Stress triggers different phases in living systems, including alarm, recovery, hardening, resistance, and exhaustion. Stress can also induce memory and is influenced by the biological clock. In addition to time, space is also an important factor in the performance of living systems under stress.
THEORETICAL AND EXPERIMENTAL PLANT PHYSIOLOGY
(2022)
Article
Biology
Robert Mendel, David Rauber, Luis A. de Souza Jr, Joao P. Papa, Christoph Palm
Summary: This paper proposes a semi-supervised approach for semantic segmentation in medical imaging research. By introducing a new paradigm of error correction, the original segmentation network is augmented to handle learning from unlabeled data. The combination of correction task and segmentation task, along with the use of a teacher-student model, leads to improved segmentation performance with limited labeled data.
COMPUTERS IN BIOLOGY AND MEDICINE
(2023)
Article
Plant Sciences
Adrya Vanessa Lira Costa, Thiago Francisco de Carvalho Oliveira, Douglas Antonio Posso, Gabriela Niemeyer Reissig, Andre Geremia Parise, Willian Silva Barros, Gustavo Maia Souza
Summary: To survive in a dynamic environment, plants have developed mechanisms to monitor and perceive the environment through electrical, hydraulic, and chemical signals. These systemic signals are induced when a stimulus is perceived and can propagate away from the stimulated site. This study evaluated the behavior of electrical and hydraulic signals in common bean plants after applying simple and combined stimuli, and observed changes in their profiles.
Article
Plant Sciences
Andre Geremia Parise, Thiago Francisco de Carvalho Oliveira, Marc-Williams Debono, Gustavo Maia Souza
Summary: Selective attention is a crucial cognitive process that enables organisms, including plants, to focus on relevant information while disregarding irrelevant stimuli. In this study, the phenomenon of attention in the parasitic plant, Cuscuta racemosa, was investigated using electrophytographic techniques. The results showed that when suitable hosts were present, the plant invested more energy in lower-frequency waves, supporting the hypothesis of attention in plants.
Article
Plant Sciences
Anderson da Rosa Feijo, Vivian Ebeling Viana, Andrisa Balbinot, Marcus Vinicius Fipke, Gustavo Maia Souza, Luciano do Amarante, Luis Antonio de Avila
Summary: This study evaluated the effect of priming rice plants with water deficit during the vegetative stage on their tolerance to heat stress during anthesis, and the contribution of atmospheric CO2 concentration. The results showed that primed plants exhibited increased yield and number of panicles, and upregulated heat shock proteins, indicating induced cross-tolerance. These findings suggest that water deficit during the vegetative stage can reduce the impact of heat stress during flowering in rice.
Article
Geochemistry & Geophysics
Daniel F. S. Santos, Joao P. Papa
Summary: This paper introduces a lightweight temporal attention network, TITAN, to address false positive and false negative alarms in change detection and reduce processing overhead. Experimental results show that the proposed approach outperforms other techniques in various evaluation metrics.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2023)
Article
Computer Science, Theory & Methods
Claudio Filipi Goncalves Dos Santos, Diego De Souza Oliveira, Leandro A. Passos, Rafael Goncalves Pires, Daniel Felipe Silva Santos, Lucas Pascotti Valem, Thierry P. Moreira, Marcos Cleison S. Santana, Mateus Roder, Joao Paulo Papa, Danilo Colombo
Summary: This article provides a survey of recent research on gait recognition using deep learning approaches, highlighting their advantages and exposing their weaknesses. It also categorizes and describes the datasets, approaches, and architectures used to tackle associated constraints.
ACM COMPUTING SURVEYS
(2023)
Article
Plant Sciences
Marcus V. Fipke, Andrisa Balbinot, Vivian E. Viana, Vinicios R. Gehrke, Magali Kemmerich, Franck E. Dayan, Gustavo M. Souza, Edinalvo R. Camargo, Luis A. Avila
Summary: This study investigated the effects of acclimatization to drought and sub-lethal doses of glyphosate on the sensitivity of Eragrostis plana to glyphosate. The findings showed that recurrent selection with drought and glyphosate stress reduced the sensitivity of the second generation to glyphosate. The upregulation of EPSPS and the ABC MRP10 transporter, as well as increased antioxidant activity, may contribute to the decreased sensitivity of the population to glyphosate.
ADVANCES IN WEED SCIENCE
(2022)
Proceedings Paper
Computer Science, Artificial Intelligence
Rafael B. M. Rodrigues, Pedro I. M. Privatto, Gustavo Jose de Sousa, Rafael P. Murari, Luis C. S. Afonso, Joao P. Papa, Daniel C. G. Pedronette, Ivan R. Guilherme, Stephan R. Perrout, Aliel F. Riente
Summary: This work introduces PetroBERT, a BERT-based model adapted for the oil and gas exploration domain in Portuguese. PetroBERT was pretrained using the Petroles corpus and a private daily drilling report corpus over BERT multilingual and BERTimbau. The proposed model was evaluated in NER and sentence classification tasks to demonstrate its potential in this domain. To the best of our knowledge, this is the first BERT-based model designed specifically for the oil and gas context.
COMPUTATIONAL PROCESSING OF THE PORTUGUESE LANGUAGE, PROPOR 2022
(2022)
Proceedings Paper
Computer Science, Artificial Intelligence
Gabriel L. Garcia, Luis C. S. Afonso, Joao P. Papa
Summary: Fake news has become a significant research topic in Natural Language Processing, and this paper proposes creating a new dataset named FakeRecogna that contains more samples, updated news, and covers important categories. Traditional classifiers and a Convolutional Neural Network are evaluated on the dataset for fake news detection.
COMPUTATIONAL PROCESSING OF THE PORTUGUESE LANGUAGE, PROPOR 2022
(2022)