113 Hits in 5.1 sec

Max-Affine Spline Insights Into Deep Network Pruning [article]

Randall Balestriero, Haoran You, Zhihan Lu, Yutong Kou, Huihong Shi, Yingyan Lin, Richard Baraniuk
2021 arXiv   pre-print
In this paper, we study the importance of pruning in Deep Networks (DNs) and the yin & yang relationship between (1) pruning highly overparametrized DNs that have been trained from random initialization  ...  pruning method, our spline pruning criteria in terms of layerwise and global pruning is on par with or even outperforms state-of-the-art pruning methods.  ...  Max-Affine Spline DNs. A key result of (Balestriero & Baraniuk, 2018) is the reformulation of current DN layers to spline operators and in particular the Max-Affine Spline Operator (MASO).  ... 
arXiv:2101.02338v2 fatcat:t5bzmjxqyjhunblhvngmhmh72i

Convolutional neural network architecture for geometric matching [article]

Ignacio Rocco, Relja Arandjelović, Josef Sivic
2017 arXiv   pre-print
We address the problem of determining correspondences between two images in agreement with a geometric model such as an affine or thin-plate spline transformation, and estimating its parameters.  ...  First, we propose a convolutional neural network architecture for geometric matching.  ...  We also provide further insights into the components of our architecture.  ... 
arXiv:1703.05593v2 fatcat:ecsen5fp7zbejbqf24yw3egxem

Sparse Bayesian Deep Learning for Dynamic System Identification [article]

Hongpeng Zhou, Chahine Ibrahim, Wei Xing Zheng, Wei Pan
2021 arXiv   pre-print
This paper proposes a sparse Bayesian treatment of deep neural networks (DNNs) for system identification.  ...  The deep neural network (DNN) models have their advantages and disadvantages.  ...  There also stands more insights into how the adopted group priors slightly change the regularization update rules to group Lasso regularizers [28] . B.  ... 
arXiv:2107.12910v1 fatcat:46umblf57zce7ju7nhpsscez7e

INSPIRE: Intensity and Spatial Information-Based Deformable Image Registration [article]

Johan Öfverstedt, Joakim Lindblad, Nataša Sladoje
2020 arXiv   pre-print
INSPIRE extends our existing symmetric registration framework based on distances combining intensity and spatial information to an elastic B-splines based transformation model.  ...  We evaluate the method on a synthetic dataset created from retinal images, consisting of thin networks of vessels, where INSPIRE exhibits excellent performance, substantially outperforming the reference  ...  be replaced with deep neural networks that directly infer the parameters of the transformation model.  ... 
arXiv:2012.07208v1 fatcat:ybpn2vrenzho7eslk4k3fcddkm

Template-Free Symbolic Performance Modeling of Analog Circuits via Canonical-Form Functions and Genetic Programming

T. McConaghy, G.G.E. Gielen
2009 IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems  
On test problems, CAFFEINE models demonstrate lower prediction error than posynomials, splines, neural networks, kriging, and support vector machines.  ...  The CAFFEINE models also had lower average prediction error than posynomials, projection-based polynomials, support vector machines, MARS splines, neural networks, and boosted neural networks.  ...  Models used include linear models [8] , [9] , [25] , posynomials [10] - [12] , polynomials [13] , [14] , [25] , splines [15] , [25] , neural networks [16] , [17] , [25] , boosted neural networks  ... 
doi:10.1109/tcad.2009.2021034 fatcat:xdktqiomdnf5lncq7tt2lo25ti

A Survey of Machine Learning Applied to Computer Architecture Design [article]

Drew D. Penney, Lizhong Chen
2019 arXiv   pre-print
This paper reviews machine learning applied system-wide to simulation and run-time optimization, and in many individual components, including memory systems, branch predictors, networks-on-chip, and GPUs  ...  Analysis also provided insight into the benefit from single-model multi-resource optimization, particularly for neural networks. Finally, recent work by Tarsa et al.  ...  Pruning applied to neural networks, either in deep Qlearning or supervised learning regression/classification, offers a method to train complex models for high accuracy, then prune for feasible implementation  ... 
arXiv:1909.12373v1 fatcat:o4nscgkjfbes7kqwmtjvvgl3oa

DeepFlow: Large Displacement Optical Flow with Deep Matching

Philippe Weinzaepfel, Jerome Revaud, Zaid Harchaoui, Cordelia Schmid
2013 2013 IEEE International Conference on Computer Vision  
The matching algorithm builds upon a multi-stage architecture with 6 layers, interleaving convolutions and max-pooling, a construction akin to deep convolutional nets.  ...  Using dense sampling, it allows to efficiently retrieve quasi-dense correspondences, and enjoys a built-in smoothing effect on descriptors matches, a valuable asset for integration into an energy minimization  ...  Our proposed matching algorithm, called deep matching, is strongly inspired by non-rigid 2D warping and deep convolutional networks [15, 28, 12] .  ... 
doi:10.1109/iccv.2013.175 dblp:conf/iccv/WeinzaepfelRHS13 fatcat:ybewiiu4w5fppcndc2bp7mks44

RANSAC-Flow: generic two-stage image alignment [article]

Xi Shen, François Darmon, Alexei A. Efros, Mathieu Aubry
2020 arXiv   pre-print
Fine alignment is learned in an unsupervised way by a deep network which optimizes a standard structural similarity metric (SSIM) between the two images, plus cycle-consistency.  ...  Our main insight is that parametric and non-parametric alignment methods have complementary strengths.  ...  thin-plate spline deformation.  ... 
arXiv:2004.01526v2 fatcat:ltpp6c4gcnfmdo3pxdt2ulqqp4

Image Matching from Handcrafted to Deep Features: A Survey

Jiayi Ma, Xingyu Jiang, Aoxiang Fan, Junjun Jiang, Junchi Yan
2020 International Journal of Computer Vision  
Finally, we conclude with the current status of image matching technologies and deliver insightful discussions and prospects for future works.  ...  Over the past decades, growing amount and diversity of methods have been proposed for image matching, particularly with the development of deep learning techniques over the recent years.  ...  This pipeline can be roughly classified into the using of classical learning and deep learning.  ... 
doi:10.1007/s11263-020-01359-2 fatcat:a2epfaolwjfm5mcrsmn7g6sd7m

The Modern Mathematics of Deep Learning [article]

Julius Berner, Philipp Grohs, Gitta Kutyniok, Philipp Petersen
2021 arXiv   pre-print
These questions concern: the outstanding generalization power of overparametrized neural networks, the role of depth in deep architectures, the apparent absence of the curse of dimensionality, the surprisingly  ...  We describe the new field of mathematical analysis of deep learning. This field emerged around a list of research questions that were not answered within the classical framework of learning theory.  ...  Below we shall discuss a couple of approaches that yield deep insights into the shape of this landscape.  ... 
arXiv:2105.04026v1 fatcat:lxnfyzr6qfasneo433inpgseia

Efficient hinging hyperplanes neural network and its application in nonlinear system identification [article]

Jun Xu, Qinghua Tao, Zhen Li, Xiangming Xi, Johan A. K. Suykens, Shuning Wang
2019 arXiv   pre-print
First the initial structure of the EHH neural network is randomly determined and the Lasso regression is used to choose the appropriate network.  ...  The construction of the EHH neural network includes 2 stages.  ...  It is worthy to note that the commonly used ReLU activation function in deep networks is a special kind of the hinge by restricting the linear function ℓ m (x) to be univariate affine.  ... 
arXiv:1905.06518v2 fatcat:iwthkypfbzbttauxir3ldvzezy

DeepMatching: Hierarchical Deformable Dense Matching [article]

Jerome Revaud, Zaid Harchaoui
2015 arXiv   pre-print
DeepMatching relies on a hierarchical, multi-layer, correlational architecture designed for matching images and was inspired by deep convolutional approaches.  ...  Relation to Deep Convolutional Neural Networks (CNNs).  ...  We then obtain: C N,p (p ) = 1 4 3 i=0 max m ∈Θi C N 2 ,p+s N,i (m ) (6) We now explain how we can break down Eq. (6) into a succession of simple operations.  ... 
arXiv:1506.07656v2 fatcat:pgez5h67njhtfgr2thphdidxre

DeepMatching: Hierarchical Deformable Dense Matching

Jerome Revaud, Philippe Weinzaepfel, Zaid Harchaoui, Cordelia Schmid
2016 International Journal of Computer Vision  
DeepMatching relies on a hierarchical, multi-layer, correlational architecture designed for matching images and was inspired by deep convolutional approaches.  ...  Relation to Deep Convolutional Neural Networks (CNNs).  ...  We then obtain: C N,p (p ) = 1 4 3 i=0 max m ∈Θi C N 2 ,p+s N,i (m ) (6) We now explain how we can break down Eq. (6) into a succession of simple operations.  ... 
doi:10.1007/s11263-016-0908-3 fatcat:umgch7rzvvcvzikmuhablkprm4

Intelligence, physics and information – the tradeoff between accuracy and simplicity in machine learning [article]

Tailin Wu
2020 arXiv   pre-print
Fourthly, to make models more robust to label noise, we introduce Rank Pruning, a robust algorithm for classification with noisy labels.  ...  This work provides deep insights about relations between the dataset, 0 and optimal representations in the Gaussian scenario, but the restriction to multivariate Gaussian datasets limits the generality  ...  Our work reveals the deep relationship between IB-Learnability and these earlier concepts and provides additional insights about what aspects of a dataset E.g., for MNIST, ( ) = 1 , where is the number  ... 
arXiv:2001.03780v2 fatcat:piduzlhoafcjhhsgthulbbhtke

Spinal Cord MRI Segmentation Techniques and Algorithms: A Survey

Sheetal Garg, S. R. Bhagyashree
2021 SN Computer Science  
Due to noise issues, deep algorithms like CNN (convolutional neural networks), RNN (recurrent neural networks) do not correctly recognize the miss-identification of the numbers in the vehicle plate.  ...  Also, in the case of deformable registration methods which are not rigid and affine the dependability is limited.  ... 
doi:10.1007/s42979-021-00618-4 fatcat:kuonhf4f7rakba6trsz67sap6m
« Previous Showing results 1 — 15 out of 113 results