Filters








1,897 Hits in 5.7 sec

Mitigating deep double descent by concatenating inputs [article]

John Chen, Qihan Wang, Anastasios Kyrillidis
2021 arXiv   pre-print
This construction empirically mitigates the double descent curve in this setting.  ...  The double descent curve is one of the most intriguing properties of deep neural networks.  ...  We artificially augment existing datasets by simply stacking every combination of inputs, and show the mitigation of the double descent curve in the deep neural network setting.  ... 
arXiv:2107.00797v1 fatcat:sqc6vxxxpbhn5cbs4ksfndcaxe

Chain-Net: Learning Deep Model for Modulation Classification Under Synthetic Channel Impairment [article]

Thien Huynh-The, Van-Sang Doan, Cam-Hao Hua, Quoc-Viet Pham, Dong-Seong Kim
2020 arXiv   pre-print
This paper proposes a robust modulation classification method by taking advantage of deep learning to capture the meaningful information of modulation signal at multi-scale feature representations.  ...  To this end, a novel architecture of convolutional neural network, namely Chain-Net, is developed with various asymmetric kernels organized in two processing flows and associated via depth-wise concatenation  ...  operation illustrated in Fig. 2(a) , two inputs having the same height and width are concatenated along the depth dimension, hence the depth size of output volume is the sum of those of two inputs.  ... 
arXiv:2009.02023v1 fatcat:odkohnlzjfa23oyu7yx653x3j4

Y-Net: A deep Convolutional Neural Network for Polyp Detection [article]

Ahmed Mohammed, Sule Yildirim, Ivar Farup, Marius Pedersen and Øistein Hovde
2018 arXiv   pre-print
To handle this problem, we propose a novel deep learning method Y-Net that consists of two encoder networks with a decoder network.  ...  Our proposed Y-Net method relies on efficient use of pre-trained and un-trained models with novel sum-skip-concatenation operations.  ...  First, frames with polyp are doubled pre-training by applying random rotation (10 • to 350 • ), zoom (1 to 1.3), translation in x,y (-10 to 10) and shear (-25 to 25) followed by centering the polyp and  ... 
arXiv:1806.01907v1 fatcat:wglp3u735nc7pcfb537q5qmqjq

Neural Edit Operations for Biological Sequences

Satoshi Koide, Keisuke Kawano, Takuro Kutsuna
2018 Neural Information Processing Systems  
The second is the use of deep CNNs with concatenations.  ...  The evolution of biological sequences, such as proteins or DNAs, is driven by the three basic edit operations: substitution, insertion, and deletion.  ...  Then, we explained that the depth and concatenation can mitigate this issue. Concatenations allow us to reuse and combine simple regular expressions (like /a/, /b/, ...).  ... 
dblp:conf/nips/KoideKK18 fatcat:6ylxv3aoijea3c5hc7g7adsrua

DeepCorn: A Semi-Supervised Deep Learning Method for High-Throughput Image-Based Corn Kernel Counting and Yield Estimation [article]

Saeed Khaki, Hieu Pham, Ye Han, Andy Kuhl, Wade Kent, Lizhi Wang
2020 arXiv   pre-print
Recent advances in machine learning, in particular deep learning, have shown promise in mitigating this bottleneck.  ...  In this paper, we propose a novel deep learning method for counting on-ear corn kernels in-field to aid in the gathering of real-time data and, ultimately, to improve decision making to maximize yield.  ...  Indeed, these works show how rapidly the combination of deep learning and plant breeding is growing and provide hope in mitigating the bottleneck facing the analysis of crop images.  ... 
arXiv:2007.10521v1 fatcat:r5w3nydjtbc45lvob6eygyzvaa

GHHT at CALCS 2018: Named Entity Recognition for Dialectal Arabic Using Neural Networks

Mohammed Attia, Younes Samih, Wolfgang Maier
2018 Proceedings of the Third Workshop on Computational Approaches to Linguistic Code-Switching  
We build a a Deep Neural Network that combines word and character-based representations in convolutional and recurrent networks with a CRF layer.  ...  It is a variant of the bi-LSTM-CRF architecture proposed by Ma and Hovy (2016) ; Lample et al. (2016) ; Huang et al. (2015) . 1 It combines a double representation of the input words by using word embeddings  ...  Training is performed using stochastic gradient descent with momentum of 0.9 and batch size equal to 150.  ... 
doi:10.18653/v1/w18-3212 dblp:conf/acl-codeswitch/AttiaSM18 fatcat:dbsp3xfxejcs3hy6k7m3q2ppmu

AOGNets: Compositional Grammatical Architectures for Deep Learning [article]

Xilai Li, Xi Song, Tianfu Wu
2019 arXiv   pre-print
Neural architectures are the foundation for improving performance of deep neural networks (DNNs).  ...  An AOG building block splits its input feature map into N groups along feature channels and then treat it as a sentence of N words.  ...  Ackowledgement This work is supported by ARO grant W911NF1810295 and DoD DURIP grant W911NF1810209.  ... 
arXiv:1711.05847v3 fatcat:b5kd3avmtbafdbpjkuvwkzy2yy

3D conditional generative adversarial networks for high-quality PET image estimation at low dose

Yan Wang, Biting Yu, Lei Wang, Chen Zu, David S. Lalush, Weili Lin, Xi Wu, Jiliu Zhou, Dinggang Shen, Luping Zhou
2018 NeuroImage  
Specifically, to render the same underlying information between the low-dose and full-dose PET images, a 3D U-net-like deep architecture which can combine hierarchical features by using skip connection  ...  Furthermore, a concatenated 3D c-GANs based progressive refinement scheme is also proposed to further improve the quality of estimated images.  ...  Acknowledgements This work was supported by National Natural Science Foundation of China (NSFC61701324) and Australian Research Council (ARC DE160100241). References  ... 
doi:10.1016/j.neuroimage.2018.03.045 pmid:29571715 pmcid:PMC6410574 fatcat:w2vzjfdtsrgkdnf3qydh7sb6wu

Automatic Mapping of Landslides by the ResU-Net

Wenwen Qi, Mengfei Wei, Wentao Yang, Chong Xu, Chao Ma
2020 Remote Sensing  
Massive landslides over large regions can be triggered by heavy rainfalls or major seismic events. Mapping regional landslides quickly is important for disaster mitigation.  ...  In this work, we proposed a deep learning approach, the ResU-Net, to map regional landslides automatically.  ...  Acknowledgments: The landslide inventory is produced by a joint effort between Wentao Yang, Chao Ma, Muyang Li, and Zhisheng Dai from Beijing Forestry University.  ... 
doi:10.3390/rs12152487 fatcat:x4pm27246jctlevezkrgx4knyy

Deep Color Transfer for Color-Plus-Mono Dual Cameras

Hae Woong Jang, Yong Ju Jung
2020 Sensors  
It removes unreliable color pixels using a reliability map computed by the binocular just-noticeable-difference model.  ...  In addition, a deep colorization network that utilizes structural information is proposed for solving the color bleeding artifact problem.  ...  Importantly, the k feature channels are concatenated with inputs [28] . The concatenated channels are again fed into the next convolution module.  ... 
doi:10.3390/s20092743 pmid:32403436 pmcid:PMC7249219 fatcat:gne5adqzgbdhbgr27fyiee4k5e

A Survey of Complex-Valued Neural Networks [article]

Joshua Bassey, Lijun Qian, Xianfang Li
2021 arXiv   pre-print
domains, where complex numbers occur either naturally or by design.  ...  Artificial neural networks (ANNs) based machine learning models and especially deep learning models have been widely applied in computer vision, signal processing, wireless communications, and many other  ...  Consequently, by representing a complex number a + ib using real numbers (a, b), the number of real parameters for each layer is doubled.  ... 
arXiv:2101.12249v1 fatcat:dntrf7fdpfgnvc3qocygmrxfny

An End-to-End Deep Neural Architecture for Optical Character Verification and Recognition in Retail Food Packaging

Fabio De Sousa Ribeiro, Liyun Gong, Francesco Caliva, Mark Swainson, Kjartan Gudmundsson, Miao Yu, Georgios Leontidis, Xujiong Ye, Stefanos Kollias
2018 2018 25th IEEE International Conference on Image Processing (ICIP)  
The proposed framework is the first to employ deep neural networks for endto-end automatic use by date recognition in retail packaging photos.  ...  In this work, an end-to-end architecture, composed of a dual deep neural network based system is proposed for automatic recognition of use by dates in food package photos.  ...  In each merging stage, the feature map from the last stage is first fed to an unpooling layer to double its size, then concatenated with the current feature map.  ... 
doi:10.1109/icip.2018.8451555 dblp:conf/icip/RibeiroGCSG0LYK18 fatcat:ijnvfwq3yrcwxfuzt7lol7rxhm

EVALUATION OF DEEP LEARNING TECHNIQUES FOR DEFORESTATION DETECTION IN THE AMAZON FOREST

M. X. Ortega, J. D. Bermudez, P. N. Happ, A. Gomes, R. Q. Feitosa
2019 ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences  
The Deep Learning-based approaches clearly outperformed the SVM baseline in our approaches, both in terms of F1-score and Overall Accuracy, with a superiority of S-CNN over EF.</p>  ...  Motivated by this scenario, this work presents an evaluation of methods for automatic deforestation detection, specifically Early Fusion (EF) Convolutional Network, Siamese Convolutional Network (S-CNN  ...  The input to EF was a tensor of a size of 15-by-15-by-16 and to S-CNN a tensor of a size of 15-by-15-by-8 in each branch and the input.  ... 
doi:10.5194/isprs-annals-iv-2-w7-121-2019 fatcat:qokcxshcincdjpbyozrdqctera

Double-Branch Network with Pyramidal Convolution and Iterative Attention for Hyperspectral Image Classification

Hao Shi, Guo Cao, Zixian Ge, Youqiang Zhang, Peng Fu
2021 Remote Sensing  
Deep-learning methods, especially convolutional neural networks (CNN), have become the first choice for hyperspectral image (HSI) classification to date.  ...  In this paper, we propose a double-branch network consisting of a novel convolution named pyramidal convolution (PyConv) and an iterative attention mechanism.  ...  By concatenating or using weighted addition, we fuse the double-branch features.  ... 
doi:10.3390/rs13071403 fatcat:r5rn5rrmtvbafdhko34y5k63di

DeepCorn: A Semi-Supervised Deep Learning Method for High-Throughput Image-Based Corn Kernel Counting and Yield Estimation [article]

Saeed Khaki, Hieu Pham, Ye Han, Andy Kuhl, Wade Kent, Lizhi Wang
2020 bioRxiv   pre-print
Recent advances in machine learning, in particular deep learning, have shown promise in mitigating this bottleneck.  ...  In this paper, we propose a novel deep learning method for counting on-ear corn kernels in-field to aid in the gathering of real-time data and, ultimately, to improve decision making to maximize yield.  ...  Additionally this work was partially supported by Syngenta.  ... 
doi:10.1101/2020.11.09.375535 fatcat:f5qtlj4qffbuljsp2fixgiahp4
« Previous Showing results 1 — 15 out of 1,897 results