共查询到19条相似文献，搜索用时 62 毫秒
1.
深度神经网络模型压缩综述 总被引：1，自引：0，他引：1
近年来;随着深度学习的飞速发展;深度神经网络受到了越来越多的关注;在许多应用领域取得了显著效果。通常;在较高的计算量下;深度神经网络的学习能力随着网络层深度的增加而不断提高;因此深度神经网络在大型数据集上的表现非常卓越。然而;由于其计算量大、存储成本高、模型复杂等特性;使得深度学习无法有效地应用于轻量级移动便携设备。因此;压缩、优化深度学习模型成为目前研究的热点。当前主要的模型压缩方法有模型裁剪、轻量级网络设计、知识蒸馏、量化、体系结构搜索等。对以上方法的性能、优缺点和最新研究成果进行了分析总结;并对未来研究方向进行了展望。 相似文献
3.
4.
5.
6.
联邦学习是一种新型的分布式机器学习方法,可以使得各客户端在不分享隐私数据的前提下共同建立共享模型。然而现有的联邦学习框架仅适用于监督学习,即默认所有客户端数据均带有标签。由于现实中标记数据难以获取,联邦学习模型训练的前提假设通常很难成立。为解决此问题,对原有联邦学习进行扩展,提出了一种基于自编码神经网络的半监督联邦学习模型ANN-SSFL,该模型允许无标记的客户端参与联邦学习。无标记数据利用自编码神经网络学习得到可被分类的潜在特征,从而在联邦学习中提供无标记数据的特征信息来作出自身贡献。在MNIST数据集上进行实验,实验结果表明,提出的ANN-SSFL模型实际可行,在监督客户端数量不变的情况下,增加无监督客户端可以提高原有联邦学习精度。 相似文献
7.
8.
9.
提出了一种基于卷积长短期记忆(LSTM)网络和卷积神经网络(CNN)的单目视觉里程计方法,命名为LSTMVO(LSTM visual odometry).LSTMVO采用无监督的端到端深度学习框架,对单目相机的6-DoF位姿以及场景深度进行同步估计.整个网络框架包含位姿估计网络以及深度估计网络,其中位姿估计网络是以端到端方式实现单目位姿估计的深度循环卷积神经网络(RCNN),由基于卷积神经网络的特征提取和基于循环神经网络(RNN)的时序建模组成,深度估计网络主要基于编码器和解码器架构生成稠密的深度图.同时本文还提出了一种新的损失函数进行网络训练,该损失函数由图像序列之间的时序损失、深度平滑度损失和前后一致性损失组成.基于KITTI数据集的实验结果表明,通过在原始单目RGB图像上进行训练,LSTMVO在位姿估计精度以及深度估计精度方面优于现有的主流单目视觉里程计方法,验证了本文提出的深度学习框架的有效性. 相似文献
10.
11.
针对现有的深度卷积神经网络往往训练平行的分类器层,很少关注类别的层次性结构,导致均衡性分类器训练难度较大的问题,提出一种结构化的深度多任务学习算法.该算法结合深度卷积神经网络与层次分类,使类别之间的结构性信息融入至深度卷积神经网络中.依托树形的类别结构设计了一个带有共享层的多分支网络结构,并使用一种关联性多任务分类器学习算法协同训练各网络分支的分类器层;为了抑制层次间的误差传播,在各分支网络的分类器层的学习过程中添加一个基于父子关系的结构化限制.采用CIFAR100和手工采集到服装数据集,在tensorflow平台上进行实验,结果表明文中算法相比于基准网络可以提高2%~4%的分类准确度. 相似文献
12.
针对多小区蜂窝网络资源分配所要求的低能耗、高速率和低延时问题,提出一种基于深度无监督学习的多小区蜂窝网络资源分配方法.首先,构建基于无监督学习的深度功率控制神经网络,通过约束处理输出优化的信道功率控制方案以最大化能量效率的期望;然后,构建基于无监督学习的深度信道分配神经网络,通过约束处理输出优化的信道分配方案,并联合前期训练好的深度功率控制神经网络拟合输出优化的信道功率,进一步优化能量效率的期望.仿真结果表明,所提出的方法在保证低计算时延的同时可获得优于其他算法的能量效率和传输速率. 相似文献
13.
基于内容的图像检索(content-based image retrieval, CBIR)是一项极具挑战的计算机视觉任务.其目标是从数据库图像中找到和查询图像包含相同实例的图像.一个典型的图像检索流程包括2步:设法从图像中提取一个合适的图像的表示向量和对这些表示向量进行最近邻搜索以找到相似的图像.其中,决定图像检索算法性能的关键在于其提取的图像表示的好坏.图像检索中使用的图像表示经历了基于手工特征和基于深度特征两大时期,每个时期又有全局特征和局部特征2个阶段.由于手工特征的表示能力有限,近年来图像检索的研究主要集中在如何利用深度特征.将以提取图像表示的不同思路为线索,回顾无监督图像检索领域的发展历程,介绍该领域的一些代表性算法,并比较这些算法在常用数据集上的性能表现,最后探讨未来的研究方向. 相似文献
14.
In this paper we present a CNN based approach for a real time 3D-hand pose estimation from the depth sequence. Prior discriminative approaches have achieved remarkable success but are facing two main challenges: Firstly, the methods are fully supervised hence require large numbers of annotated training data to extract the dynamic information from a hand representation. Secondly, unreliable hand detectors based on strong assumptions or a weak detector which often fail in several situations like complex environment and multiple hands. In contrast to these methods, this paper presents an approach that can be considered as semi-supervised by performing predictive coding of image sequences of hand poses in order to capture latent features underlying a given image without supervision. The hand is modelled using a novel latent tree dependency model (LDTM) which transforms internal joint location to an explicit representation. Then the modeled hand topology is integrated with the pose estimator using data dependent method to jointly learn latent variables of the posterior pose appearance and the pose configuration respectively. Finally, an unsupervised error term which is a part of the recurrent architecture ensures smooth estimations of the final pose. Experiments on three challenging public datasets, ICVL, MSRA, and NYU demonstrate the significant performance of the proposed method which is comparable or better than state-of-the-art approaches. 相似文献
15.
Daniel Peralta Isaac Triguero Salvador García Yvan Saeys Jose M. Benitez Francisco Herrera 《国际智能系统杂志》2018,33(1):213-230
Fingerprint classification is one of the most common approaches to accelerate the identification in large databases of fingerprints. Fingerprints are grouped into disjoint classes, so that an input fingerprint is compared only with those belonging to the predicted class, reducing the penetration rate of the search. The classification procedure usually starts by the extraction of features from the fingerprint image, frequently based on visual characteristics. In this work, we propose an approach to fingerprint classification using convolutional neural networks, which avoid the necessity of an explicit feature extraction process by incorporating the image processing within the training of the classifier. Furthermore, such an approach is able to predict a class even for low‐quality fingerprints that are rejected by commonly used algorithms, such as FingerCode. The study gives special importance to the robustness of the classification for different impressions of the same fingerprint, aiming to minimize the penetration in the database. In our experiments, convolutional neural networks yielded better accuracy and penetration rate than state‐of‐the‐art classifiers based on explicit feature extraction. The tested networks also improved on the runtime, as a result of the joint optimization of both feature extraction and classification. 相似文献
16.
This paper presents a learning rule, CBA, to develop oriented receptive fields similar to those founded in cat striate cortex.
The inherent complexity of the development of selectivity in visual cortex has led most authors to test their models by using
a restricted input environment. Only recently, some learning rules (the PCA and the BCM rules) have been studied in a realistic
visual environment. For these rules, which are based upon Hebbian learning, single neuron models have been proposed in order
to get a better understanding of their properties and dynamics. These models suffered from unbounded growing of synaptic strength,
which is remedied by a normalization process. However, normalization seems biologically implausible, given the non-local nature
of this process. A detailed stability analysis of the proposed rule proves that the CBA attains a stable state without any
need for normalization. Also, a comparison among the results achieved in different types of visual environments by the PCA,
the BCM and the CBA rules is provided. The final results show that the CBA rule is appropriate for studying the biological
process of receptive field formation and its application in image processing and artificial vision tasks. 相似文献
17.
Michael S. Gashler Michael R. Smith Richard Morris Tony Martinez 《Computational Intelligence》2016,32(2):196-215
Many data mining and data analysis techniques operate on dense matrices or complete tables of data. Real‐world data sets, however, often contain unknown values. Even many classification algorithms that are designed to operate with missing values still exhibit deteriorated accuracy. One approach to handling missing values is to fill in (impute) the missing values. In this article, we present a technique for unsupervised learning called unsupervised backpropagation (UBP), which trains a multilayer perceptron to fit to the manifold sampled by a set of observed point vectors. We evaluate UBP with the task of imputing missing values in data sets and show that UBP is able to predict missing values with significantly lower sum of squared error than other collaborative filtering and imputation techniques. We also demonstrate with 24 data sets and nine supervised learning algorithms that classification accuracy is usually higher when randomly withheld values are imputed using UBP, rather than with other methods. 相似文献
18.
生物医学成像领域的迅速发展引起相关图像信息的爆炸式增长,对其图像进行人工智能辅助分析日益成为科学研究、临床应用、即时诊断等领域的迫切需求.近年来深度学习,尤其是卷积神经网络在生物医学图像分析领域取得广泛应用,在生物医学图像的信息提取,包括细胞分类、检测,生理及病理图像的分割、检测等领域发挥日益重要的作用.介绍了深度学习... 相似文献
19.
In our contribution we demonstrate, that orientation (OR) and ocular dominance (OD) cannot develop simultaneously in linear correlation-based learning (CBL) models, because OR and OD occupy separate domains in parameter space. We then analyse the conditions under which waves of spontaneous activity — as have been observed in the developing retina — may give rise at least to OR or OD. We find that in linear CBL models there must be subcortical convergence of ON/OFF as well as left eye/right eye pathways. We present a biologically plausible scenario in which the subcortical connectivity patterns resemble a competitive neural network and argue that OR and OD can robustly emerge, if a two-stage developmental process is assumed. 相似文献