人工智能培训

搜索

论文代码开源:点击率预测的深度兴趣演化网络(Deep Interest Evolution Network for Click-Through Rate Pre

[复制链接]
admin 发表于 2018-9-15 10:22:50 | 显示全部楼层 |阅读模式
admin 2018-9-15 10:22:50 1108 0 显示全部楼层
人工智能论文代码开源:点击率预测的深度兴趣演化网络(Deep Interest Evolution Network for Click-Through Rate Prediction)请注意该人工智能论文代码开源在github,大部分是python写的,框架可能是tensorflow或者pytorch。时间序列分类(TSC)是印度采矿的一个重要且具有挑战性的问题。随着时间序列数据可用性的增加,已经提出了数百种TSC算法。在这些方法中,只有少数人考虑过深度神经网络(DNN)来执行此任务。令人惊讶的是,深度学习在过去几年中已经看到非常成功的应用。 DNNshave确实彻底改变了计算机视觉领域,特别是随着新的更深层次的体系结构(如残余和卷积神经网络)的出现。除了图像之外,还可以使用DNN处理诸如文本和音频的顺序数据,以达到用于文档分类和语音识别的最先进性能。在本文中,我们通过对TSC最新DNN架构的实证研究,研究了TSC深度学习算法的当前表现。我们在TSN的统一分类标准下对各个时间序列域中最成功的深度学习应用进行了概述。我们还为TSC社区提供了一个开源的深度学习框架,我们实施了比较方法,并在单变量TSCbenchmark(UCR存档)和12个多变量时间序列数据集上进行了评估。通过在97个时间序列数据集上训练8,730个深度学习模型,我们提出了迄今为止针对TSC的DNN的详尽研究。
Time Series Classification (TSC) is an important and challenging problem indata mining.With the increase of time series data availability, hundreds ofTSC algorithms have been proposed.Among these methods, only a few haveconsidered Deep Neural Networks (DNNs) to perform this task.This is surprisingas deep learning has seen very successful applications in the last years.DNNshave indeed revolutionized the field of computer vision especially with theadvent of novel deeper architectures such as Residual and Convolutional NeuralNetworks.Apart from images, sequential data such as text and audio can also beprocessed with DNNs to reach state of the art performance for documentclassification and speech recognition.In this article, we study the currentstate of the art performance of deep learning algorithms for TSC by presentingan empirical study of the most recent DNN architectures for TSC.We give anoverview of the most successful deep learning applications in various timeseries domains under a unified taxonomy of DNNs for TSC.We also provide anopen source deep learning framework to the TSC community where we implementedeach of the compared approaches and evaluated them on a univariate TSCbenchmark (the UCR archive) and 12 multivariate time series datasets.Bytraining 8,730 deep learning models on 97 time series datasets, we propose themost exhaustive study of DNNs for TSC to date.论文代码开源:点击率预测的深度兴趣演化网络(Deep Interest Evolution Network for Click-Through Rate Prediction) X79yBO9NA5W5Yo7w.jpg
URL地址:https://arxiv.org/abs/1809.04356v1     ----pdf下载地址:https://arxiv.org/pdf/1809.04356v1    ----         ----github下载地址:https://github.com/hfawaz/dl-4-tsc    ----    论文代码开源:点击率预测的深度兴趣演化网络(Deep Interest Evolution Network for Click-Through Rate Prediction)请注意该人工智能论文代码开源在github,大部分是python写的,框架可能是tensorflow或者pytorch,keras,至于具体是哪一个没有完全测试。
回复

使用道具 举报

您需要登录后才可以回帖 登录 | 立即注册

本版积分规则 返回列表 发新帖

admin当前离线
管理员

查看:1108 | 回复:0

快速回复 返回顶部 返回列表