Filters








7,310 Hits in 7.1 sec

k-Sparse Autoencoders [article]

Alireza Makhzani, Brendan Frey
2014 arXiv   pre-print
Recently, it has been observed that when representations are learnt in a way that encourages sparsity, improved performance is obtained on classification tasks.  ...  are simple to train and the encoding stage is very fast, making them well-suited to large problem sizes, where conventional sparse coding algorithms cannot be applied.  ...  However, these features could be used for pre-training deep neural nets.  ... 
arXiv:1312.5663v2 fatcat:dyhkejyisjapbk4ib4hpydbmtq

Deep learning applications and challenges in big data analytics

Maryam M Najafabadi, Flavio Villanustre, Taghi M Khoshgoftaar, Naeem Seliya, Randall Wald, Edin Muharemagic
2015 Journal of Big Data  
We also investigate some aspects of Deep Learning research that need further exploration to incorporate specific challenges introduced by Big Data Analytics, including streaming data, high-dimensional  ...  Complex abstractions are learnt at a given level based on relatively simpler abstractions formulated in the preceding level in the hierarchy.  ...  For example, Google and Stanford formulated a very large deep neural network that was able to learn very high-level features, such as face detection or cat detection from scratch (without any priors) by  ... 
doi:10.1186/s40537-014-0007-7 fatcat:65mi6dnv5rg6poesotupqbsm7y

Understanding Internal Semantics Of Deep Learning Models For Electronic Music

Minz Sanghee Won
2017 Zenodo  
With conventional approaches and proposed methods, I investigate latent semantics learnt in hidden layers of deep learning models, particularly Convolutional Neural Networks (CNNs).  ...  The experimental result reports latent semantics of learnt kernels from pre-trained CNNs for the electronic music genre classification, which was not mainly explored in a deep learning research.  ...  In the deeper layer, the network learnt high-level features.  ... 
doi:10.5281/zenodo.1100967 fatcat:fewtmbdobbepdbslgt2phdie5y

Radar HRRP recognition based on CNN

Jia Song, Yanhua Wang, Wei Chen, Yang Li, Junfu Wang
2019 The Journal of Engineering  
In this study, ground target recognition based on one-dimensional convolutional neural network (CNN) is studied by exploiting the targets' high-resolution range profiles (HRRPs).  ...  Contrary to conventional methods which need feature extraction artificially, CNN can automatically discover features for classification.  ...  Acknowledgments This work was supported in part by the National Natural Science Foundation of China (Grant no. 61701026), the Chang Jiang Scholars Programme (Grant no.  ... 
doi:10.1049/joe.2019.0725 fatcat:rqcmpv5yu5ah5k3kgi5mlmrz3u

Multi-Phase Feature Representation Learning for Neurodegenerative Disease Diagnosis [chapter]

Siqi Liu, Sidong Liu, Weidong Cai, Sonia Pujol, Ron Kikinis, David Dagan Feng
2015 Lecture Notes in Computer Science  
MPFR learns high-level neuroimaging features by extracting the associations between the low-dimensional biomarkers and the high-dimensional neuroimaging features with a deep neural network.  ...  The proposed approach outperformed the original neural network in both binary and ternary Alzheimer's disease classification tasks.  ...  Acknowledgements This work was supported in part by the ARC, AADRF, NA-MIC (NIH U54EB005149), and NAC (NIH P41EB015902).  ... 
doi:10.1007/978-3-319-14803-8_27 fatcat:6r46rorxnfhbjdq4by4uz2xgga

HAR-Net:Fusing Deep Representation and Hand-crafted Features for Human Activity Recognition [article]

Mingtao Dong, Jindong Han
2018 arXiv   pre-print
Conventional HAR based on Support Vector Machine relies on subjective manually extracted features.  ...  The HAR-Net fusing the hand-crafted features and high-level features extracted from convolutional network to make prediction.  ...  The HAR-Net combined the Hand-crafted features with the learnt features by the deep neural network to provide more comprehensive perspectives when making predictions.  ... 
arXiv:1810.10929v1 fatcat:gt2erzxob5hizdjeldnbxxf5cq

Explicitising The Implicit Intrepretability of Deep Neural Networks Via Duality [article]

Chandrashekar Lakshminarayanan, Amit Vikram Singh, Arun Rajkumar
2022 arXiv   pre-print
Recent work by Lakshminarayanan and Singh [2020] provided a dual view for fully connected deep neural networks (DNNs) with rectified linear units (ReLU).  ...  These results motivate a novel class of networks which we call deep linearly gated networks (DLGNs).  ...  The gating network is also called as the feature network since it realises the neural path features, and the weight network is also called as the value network since it realises the neural path value.  ... 
arXiv:2203.16455v1 fatcat:sevasqw67jfcpifb2pzlgz3w3i

Exploiting Cascaded Ensemble of Features for the Detection of Tuberculosis Using Chest Radiographs

A F M Saif, Tamjid Imtiaz, Celia Shahnaz, Wei Ping Zhu, M. O. Ahmad
2021 IEEE Access  
Considering all these facts, a cascaded ensembling method is proposed that combines both the hand-engineered and the deep learning-based features for the Tuberculosis detection task.  ...  However, at present, deep learning (DL) gains popularity in many computer vision tasks because of their better performance in comparison to the traditional manual feature extraction based machine learning  ...  In Table. 3, performance of the individual transfer learnt neural network features are summarized.  ... 
doi:10.1109/access.2021.3102077 fatcat:34c3rfrirvc5jpudneijt5rxr4

Design of "Deep Learning Controller"

Koksal Erenturk
2018 International Journal of Engineering and Applied Sciences (IJEAS)  
In the meantime, recent advances in deep learning, encompassing neural networks, hierarchical probabilistic models, and a variety of unsupervised and supervised feature learning algorithms, have brought  ...  Deep neural networks (DNN) serve as function approximators and are used to learn the control policies. Once the DNN trained, control actions can be achieved at the output of the learned network.  ...  Using these features, large and complex problems that could not be solved with conventional neural networks can be resolved by deep learning algorithms.  ... 
doi:10.31873/ijeas.5.10.31 fatcat:sy5a2kkt7nbljgauq2dnbfjbdm

An Effective Model for Detection of Dysfunctionality in Heart Based on Iridology using Deep Neural Networks

2020 VOLUME-8 ISSUE-10, AUGUST 2019, REGULAR ISSUE  
convolutional neural network.  ...  this article, we examine the heart dysfunctionality through a chain of steps which are localization of iris, segmentation of iris, ROI extraction, histogram equalization of ROI and classification using deep  ...  The filters that are hand-engineered in conventional methods for feature extraction are learnt by the neural networks on their own which is a major advantage of deep learning over conventional methods.  ... 
doi:10.35940/ijitee.e2888.039520 fatcat:i3iba434yvc5hnaalbcznag4dm

Still Image-based Human Activity Recognition with Deep Representations and Residual Learning

Ahsan Raza Siyal, Zuhaibuddin Bhutto, Syed Muhammad, Azhar Iqbal, Faraz Mehmood, Ayaz Hussain, Saleem Ahmed
2020 International Journal of Advanced Computer Science and Applications  
In this paper, a novel method is presented for human activity recognition based on pretrained Convolutional Neural Network (CNN) model utilized as feature extractor and deep representations are followed  ...  by Support Vector Machine (SVM) classifier for action recognition.  ...  Transfer learning is the deep learning technique to set a deep neural network using features learnt for a source problem ( ) , TS DS and the same network can be fine-tuned and employed for target task  ... 
doi:10.14569/ijacsa.2020.0110561 fatcat:u7jukwoznbh2bev2xpbgftvgui

Artificial Intelligence – State of Art Convolution Neural Network Architectures in a Nutshell

2019 VOLUME-8 ISSUE-10, AUGUST 2019, REGULAR ISSUE  
The Convolution neural network (CNN) is one of best deep architecture suitable to handle variety of inputs.  ...  the human brain function by understanding how each part of the brain works.  ...  In such systems the CNN predicts the neural activity from features learnt from input sound stimuli.  ... 
doi:10.35940/ijitee.k1257.09811s19 fatcat:jedj52arj5etfeepmnhd4aurqi

EEG Based Eye State Classification using Deep Belief Network and Stacked AutoEncoder

Sanam Narejo, Eros Pasero, Farzana Kulsoom
2016 International Journal of Electrical and Computer Engineering (IJECE)  
Recent studies reflect that Deep Neural Networks are trending state of the art Machine learning approaches.  ...  Therefore, the current work presents the implementation of Deep Belief Network (DBN) and Stacked AutoEncoders (SAE) as Classifiers with encouraging performance accuracy.  ...  ACKNOWLEDGEMENTS This research activity was partly funded by Italian MIUR OPLON project and supported by the Politecnico of Turin NEC NEURONICA laboratory.  ... 
doi:10.11591/ijece.v6i6.12967 fatcat:l5fxoxzooveytlhit2xgnvpc7e

Automatic Feature Learning Method for Detection of Retinal Landmarks

Baidaa Al-Bander, Waleed Al-Nuaimy, Majid A. Al-Taee, Ali Al-Ataby, Yalin Zheng
2016 2016 9th International Conference on Developments in eSystems Engineering (DeSE)  
The proposed method, which is based on deep convolutional neural networks (CNN) does not depend the visual appearance or anatomical features of the retinal landmarks.  ...  Performance of the network is evaluated using Root Mean Square Error (RMSE). The developed feature learning approach presents a promising system for retinal landmark detection.  ...  features are learnt by deep learning algorithms directly from the raw input data.  ... 
doi:10.1109/dese.2016.4 fatcat:mzgjghvyyjhrpmh53u37zvnkda

Learning Image-based Representations for Heart Sound Classification

Zhao Ren, Nicholas Cummins, Vedhas Pandit, Jing Han, Kun Qian, Björn Schuller
2018 Proceedings of the 2018 International Conference on Digital Health - DH '18  
When compared to a baseline accuracy of 46.9 %, gained using conventional audio processing features and a support vector machine, this is a significant relative improvement of 19.8 % (p < .001 by one-tailed  ...  Deep representations are then extracted from a fully connected layer of each network and classification is achieved by a static classifier.  ...  than a 'fixed' conventional feature set.  ... 
doi:10.1145/3194658.3194671 dblp:conf/ehealth/RenCPH0S18 fatcat:zh74eahd6ffhrahnf7grr7m7mu
« Previous Showing results 1 — 15 out of 7,310 results