Filters








29,824 Hits in 6.7 sec

Contraction Mapping of Feature Norms for Classifier Learning on the Data with Different Quality [article]

Weihua Liu, Xiabi Liu, Murong Wang, Ling Ma
2020 arXiv   pre-print
However, the data for training image classifiers usually has different quality. Ignoring such problem, the correct classification of low quality data is hard to be solved.  ...  Based on this finding, we propose a contraction mapping function to compress the range of feature norms of training images according to their quality and embed this contraction mapping function into softmax  ...  to effectively deal with the problem of the difference of data in quality and thus can bring significant and stable performance boost to the learning based on the softmax losses.  ... 
arXiv:2007.13406v2 fatcat:ksedckeicjestikbthaluk3bq4

Higher Order Contractive Auto-Encoder [chapter]

Salah Rifai, Grégoire Mesnil, Pascal Vincent, Xavier Muller, Yoshua Bengio, Yann Dauphin, Xavier Glorot
2011 Lecture Notes in Computer Science  
We explicitly encourage the latent representation to contract the input space by regularizing the norm of the Jacobian (analytically) and the Hessian (stochastically) of the encoder's output with respect  ...  While the penalty on the Jacobian's norm ensures robustness to tiny corruption of samples in the input space, constraining the norm of the Hessian extends this robustness when moving further away from  ...  Then, to roughly reduce the dimension, features are sum-pooled together over quadrants of the feature maps. So the input dimension of the linear classifier is equal to 4n h .  ... 
doi:10.1007/978-3-642-23783-6_41 fatcat:c2bgqwwlpbd4rk2ernl7vf5eme

TSInsight: A local-global attribution framework for interpretability in time-series data [article]

Shoaib Ahmed Siddiqui, Dominique Mercier, Andreas Dengel, Sheraz Ahmed
2020 arXiv   pre-print
We approach the problem of interpretability in a novel way by proposing TSInsight where we attach an auto-encoder to the classifier with a sparsity-inducing norm on its output and fine-tune it based on  ...  TSInsight learns to preserve features that are important for prediction by the classifier and suppresses those that are irrelevant i.e. serves as a feature attribution method to boost interpretability.  ...  We would like to further investigate the automated selection of hyperparameters in the future which is primitive for the wide-scale applicability of TSInsight along with its impact on adversarial robustness  ... 
arXiv:2004.02958v1 fatcat:xeutmwgkgjggxhwcpphwn66cwi

TSInsight: A Local-global Attribution Framework for Interpretability in Time Series Data

Shoaib Ahmed Siddiqui, Dominique Mercier, Andreas Dengel, Sheraz Ahmed
2021 Sensors  
We approach the problem of interpretability in a novel way by proposing TSInsight, where we attach an auto-encoder to the classifier with a sparsity-inducing norm on its output and fine-tune it based on  ...  TSInsight learns to preserve features that are important for prediction by the classifier and suppresses those that are irrelevant, i.e., serves as a feature attribution method to boost the interpretability  ...  Data Availability Statement: Not applicable. Conflicts of Interest: The authors declare no conflict of interest.  ... 
doi:10.3390/s21217373 pmid:34770678 pmcid:PMC8587116 fatcat:jn6ujhwhqber3g6qab5sr4b4ga

Evaluating the Robustness of Defense Mechanisms based on AutoEncoder Reconstructions against Carlini-Wagner Adversarial Attacks

Petru Hlihor, Riccardo Volpi, Luigi Malagò
2020 Proceedings of the Northern Lights Deep Learning Workshop  
on adversarial examples generated with the Carlini-Wagner attack, in a white-box scenario and on the stacked system composed by the autoencoder and the classifier.  ...  Adversarial Examples represent a serious problem affecting the security of machine learning systems.  ...  2014-2020, Action 1.1.4, project ID P 37 714, contract no. 136/27.09.2016.  ... 
doi:10.7557/18.5173 fatcat:hvjupjkjafebjmxtjmzwenz2iu

Comparing U-Net Based Models for Denoising Color Images

Rina Komatsu, Tad Gonsalves
2020 AI  
This study deals with the design and training of a generalized deep learning denoising model that can remove five different kinds of noise from any digital image: Gaussian noise, salt-and-pepper noise,  ...  Although the kinds of digital noise are varied, current denoising studies focus on denoising only a single and specific kind of noise using a devoted deep-learning model.  ...  When up-sampling feature maps in the U-Net, the outputs from the previous deconvolution layer in the expanding path are concatenated with the feature maps obtained through the contracting path.  ... 
doi:10.3390/ai1040029 fatcat:4qns4lgtvvgkvkyjk6xwia2gzi

Unsupervised Learning of Semantics of Object Detections for Scene Categorization [chapter]

Grégoire Mesnil, Salah Rifai, Antoine Bordes, Xavier Glorot, Yoshua Bengio, Pascal Vincent
2014 Advances in Intelligent Systems and Computing  
We validate the quality of our models in an extensive set of experiments in which several setups of the sequential feature extraction process are evaluated on benchmarks for scene classification [21, 23  ...  Standard approaches build an intermediate representation of the global image and learn classifiers on it.  ...  Codes for the experiments have been implemented using Theano [4] Machine Learning library.  ... 
doi:10.1007/978-3-319-12610-4_13 fatcat:a36rlagtizg4levkrxozjaiehq

Applying advanced machine learning models to classify electro-physiological activity of human brain for use in biometric identification [article]

Iaroslav Omelianenko
2017 arXiv   pre-print
After that, we continue with a description of the advanced machine learning pipeline model that can perform confident classification of the collected EEG data in order to (a) reliably distinguish signal  ...  All this in combination opens up broad prospects for the development of new types of consumer devices, [e.g.] based on virtual reality helmets or augmented reality glasses where EEG sensor can be easily  ...  We believe that further improvement of the method can provide sufficiently reliable means for use even for user identification based on the collected EEG data.  ... 
arXiv:1708.01167v1 fatcat:3p5kv77xqrdmxoetds6fkfghti

Manifold regularization with GANs for semi-supervised learning [article]

Bruno Lecouat, Chuan-Sheng Foo, Houssam Zenati, Vijay Chandrasekhar
2018 arXiv   pre-print
When incorporated into the semi-supervised feature-matching GAN we achieve state-of-the-art results for GAN-based semi-supervised learning on CIFAR-10 and SVHN benchmarks, with a method that is significantly  ...  We leverage this property to perform manifold regularization by approximating a variant of the Laplacian norm using a Monte Carlo approximation that is easily computed with the GAN.  ...  In order to quantify the importance of the first step of manifold learning, we also compared the performance of classifiers when regularized using GANs of differing quality (as assessed by the quality  ... 
arXiv:1807.04307v1 fatcat:deaj5qucafdd3eapcy4tzklicq

A pan-cancer somatic mutation embedding using autoencoders

Martin Palazzo, Pierre Beauseroy, Patricio Yankilevich
2019 BMC Bioinformatics  
Kernel learning with hierarchical cluster analysis are used to assess the quality of the learned somatic mutation embedding, on which support vector machine models are used to accurately classify tumor  ...  The increasing availability of tumor data allows to research the complexity of cancer disease with machine learning methods.  ...  Acknowledgements The authors would like to thank the International Cancer Genome Consortium who have provided the data.  ... 
doi:10.1186/s12859-019-3298-z pmid:31829157 pmcid:PMC6907172 fatcat:dzw7gk7ubjhbrgu6t5baocjwi4

What Can Artificial Intelligence Offer Coral Reef Managers?

Sarah M. Hamylton, Zhexuan Zhou, Lei Wang
2020 Frontiers in Marine Science  
Working with covariate data and desired results, a machine learning classifier reliably turned one into the other and applied at a global scale using Google Earth Engine to provide a repository of high  ...  Taking a fundamentally different approach to most commercial image interpretation softwares, machine learning algorithms work with data and desired results to generate a model that turns one into the other  ... 
doi:10.3389/fmars.2020.603829 fatcat:ttb6yyk3xvdqvgtjyjyuvsrghq

Abnormality detection and intelligent severity assessment of human chest computed tomography scans using deep learning: a case study on SARS-COV-2 assessment

Mohamed Ramzy Ibrahim, Sherin M. Youssef, Karma M. Fathalla
2021 Journal of Ambient Intelligence and Humanized Computing  
In this paper, a novel computer aided framework (COV-CAF) is proposed for classifying the severity degree of the infection from 3D Chest Volumes.  ...  COV-CAF fuses traditional and deep learning approaches. The proposed COV-CAF consists of two phases: the preparatory phase and the feature analysis and classification phase.  ...  Ethics approval Not Applicable Data are obtained from publicly available datasets https:// mosmed. ai/ en/ & www. kaggle. com/ plame nedua rdo/ sarsc ov2-ctscan-datas et.  ... 
doi:10.1007/s12652-021-03282-x pmid:34055098 pmcid:PMC8147594 fatcat:dx637yujpzfpnppiaoxqmgocei

Segmenting two-dimensional structures with strided tensor networks [article]

Raghavendra Selvan, Erik B Dam, Jens Petersen
2021 arXiv   pre-print
Additionally, based on the experiments we discuss the feasibility of using fully linear models for segmentation tasks.  ...  More recently, supervised learning has been attempted with tensor networks, primarily focused on tasks such as image classification.  ...  Further, the local feature maps are constrained to have unit norm so that the global feature map in the next step also has unit norm.  ... 
arXiv:2102.06900v1 fatcat:yc6m3pa6mzdsdpbgaemrnw3gmm

FREA-Unet: Frequency-aware U-net for Modality Transfer [article]

Hajar Emami, Qiong Liu, Ming Dong
2020 arXiv   pre-print
Our frequency-aware attention Unet computes the attention scores for feature maps in low/high frequency layers and use it to help the model focus more on the most important regions, leading to more realistic  ...  Specifically, we incorporate attention mechanism into different U-net layers responsible for estimating low/high frequency scales of the image.  ...  The model is trained with ADAM optimizer with an initial learning rate of 0.0002 and with a batch size of 1. We trained the model for 200 epochs on a NVIDIA GTX 1080 Ti GPU.  ... 
arXiv:2012.15397v1 fatcat:nuitbwklqbdmzja5ueccaghrja

Deep Neural Framework with Visual Attention and Global Context for Predicting Image Aesthetics

Yifei Xu, Nuo Zhang, Pingping Wei, Genan Sang, Li Li, Feng Yuan
2020 IEEE Access  
Then, the 1×1×N feature map is duplicated M × M copies and then concatenated with the original M × M × N feature map, resulting a M × M × 2N concatenated feature map.  ...  Thought we solve the over-fitting problem caused by insufficient data with pretrained models, the burden of collecting large-scale supervised data for industrial needs is still challenging.  ... 
doi:10.1109/access.2020.3015060 fatcat:3etz464ha5hbvcbe66ufvpcymy
« Previous Showing results 1 — 15 out of 29,824 results