Set Aggregation Network as a Trainable Pooling Layer

机译：将聚合网络设置为可训练的池层

获取原文

页面导航

摘要
著录项
相似文献
相关主题

摘要

Global pooling, such as max- or sum-pooling, is one of the key ingredients in deep neural networks used for processing images, texts, graphs and other types of structured data. Based on the recent DeepSets architecture proposed by Zaheer et al. (NIPS 2017), we introduce a Set Aggregation Network (SAN) as an alternative global pooling layer. In contrast to typical pooling operators, SAN allows to embed a given set of features to a vector representation of arbitrary size. We show that by adjusting the size of embedding, SAN is capable of preserving the whole information from the input. In experiments, we demonstrate that replacing global pooling layer by SAN leads to the improvement of classification accuracy. Moreover, it is less prone to overfitting and can be used as a regularizer.

机译：全局池（例如最大池或总池）是深度神经网络中用于处理图像，文本，图形和其他类型的结构化数据的关键要素之一。基于Zaheer等人最近提出的DeepSets体系结构。（NIPS 2017），我们引入了集合聚合网络（SAN）作为替代的全局池化层。与典型的池化运算符相比，SAN允许将给定的功能集嵌入到任意大小的矢量表示中。我们表明，通过调整嵌入的大小，SAN能够保留来自输入的全部信息。在实验中，我们证明了用SAN替换全局池层可以提高分类准确性。此外，它不太容易过度拟合，可以用作正则化器。

著录项

来源
《International conference on neural information processing;Annual conference of Asia-Pacific Neural Network Society》|2019年|419-431|共13页
会议地点
作者
Lukasz Maziarka; Marek Smieja; Aleksandra Nowak; Jacek Tabor; Lukasz Struski; Przemyslaw Spurek;
展开▼
作者单位

展开▼
会议组织
原文格式 PDF
正文语种
中图分类
关键词
Global pooling; Structured data; Representation learning; Convolutional neural networks; Set processing; Image processing;

机译：全局池;结构化数据;表征学习;卷积神经网络设置处理;图像处理;

相似文献

外文文献
中文文献
专利

1. Chaotic fractal walk trainer for sonar data set classification using multi-layer perceptron neural network and its hardware implementation [J] . Khishe M., Mosavi M. R., Moridi A. Applied Acoustics . 2018,第AUGa期

机译：多层感知器神经网络的声纳数据集分类混沌分形步行训练器及其硬件实现
2. Learning Precise Spike Train–to–Spike Train Transformations in Multilayer Feedforward Neuronal Networks [J] . Arunava Banerjee Neural computation . 2016,第5期

机译：在多层前馈神经元网络中学习精确的尖峰火车到尖峰火车转换
3. The consequences of the aggregation of detritus pools in ecological networks [J] . Allesina S, Bondavalli C, Scharler UM, Ecological Modelling . 2005,第1a2期

机译：生态网络中碎屑池聚集的后果
4. Set Aggregation Network as a Trainable Pooling Layer [C] . Lukasz Maziarka, Marek Smieja, Aleksandra Nowak, International Conference on Neural Information Processing . 2019

机译：将聚合网络设置为培训池层
5. Rank Aggregation Methods For Consensus Ranking in Multilayer Networks [D] . Braun, Niklas 2019

机译：多层网络中共识排序的排序聚合方法
6. Modelling spatially-resolved diffuse reflectance spectra of a multi-layered skin model by artificial neural networks trained with Monte Carlo simulations [O] . Sheng-Yang Tsui, Chiao-Yi Wang, Tsan-Hsueh Huang, 2018

机译：通过蒙特卡洛模拟训练的人工神经网络对多层皮肤模型的空间分辨漫反射光谱建模
7. Table 5: Accuracy of SVM classifier on DBAP features derived from pre-trained LeNet with DBAP layer. The DBAP features show better classification results than the MaxPool features in LeNet. The fully connected (FC) layers of LeNet with DBAP also tend to show better discrimination ability as compared to FC layer features extracted from regular LeNet on all benchmark data sets. [O] . -1

机译：表5：SVM分类器对具有DBAP层的预先训练的Lenet派生的DBAP特征的精度。 DBAP功能显示比LENET中的MAXPOOL功能更好的分类结果。与DBAP的完全连接（FC）LENET的层也倾向于显示与在所有基准数据集上从常规LENET提取的FC层特征相比显示出更好的辨别能力。
8. A General Class of Layered, Trainable, Threshold Logic Networks for Pattern Classification [R] . Francalangia, J. L. 1969

机译：一类用于模式分类的分层可训练门限逻辑网络

Set Aggregation Network as a Trainable Pooling Layer

摘要

著录项

相似文献

相关主题

期刊订阅