imported>JimGrassroot：使用HotCat已添加Category:开源人工智能

2025-06-08T03:22:55Z

新页面

{{TA|G1=IT}}
{{Infobox software
| name = Deeplearning4j
| logo =
| screenshot =
| caption =
| collapsible =
| author = [[Adam Gibson]]
| developer = [https://github.com/SkymindIO/deeplearning4j/graphs/contributors 众多]
| released =
| programming language = [[Java]], [[Scala]], [[CUDA]], [[C语言|C]]
| operating system = [[Linux]], [[macOS]], [[Windows]], [[Android]]
| platform = [[跨平台]]
| size =
| language =
| genre = [[自然语言处理]], [[深度学习]], [[机器视觉]], [[人工智能]]
| license = [[Apache许可证|Apache许可证2.0]]
| website = {{URL|http://deeplearning4j.org}}
}}

'''Deeplearning4j'''是为[[Java]]和[[Java虚拟机]]<ref name="wired">{{cite web|first=Cade|last=Metz|title=The Mission to Bring Google's AI to the Rest of the World|work=[[Wired.com]]|date=2014-06-02|url=http://www.wired.com/2014/06/skymind-deep-learning/|accessdate=2014-06-28|archive-date=2014-06-29|archive-url=https://web.archive.org/web/20140629065833/http://www.wired.com/2014/06/skymind-deep-learning/|dead-url=no}}</ref><ref>{{cite web|url=http://www.businessweek.com/articles/2014-06-03/teaching-smaller-companies-how-to-probe-deep-learning-on-their-own|title=Deep Learning for (Some of) the People|last=Vance|first=Ashlee|work=[[Bloomberg Businessweek]]|date=2014-06-03|accessdate=2014-06-28|archive-date=2014-06-25|archive-url=https://web.archive.org/web/20140625071114/http://www.businessweek.com/articles/2014-06-03/teaching-smaller-companies-how-to-probe-deep-learning-on-their-own|dead-url=no}}</ref>编写的[[开源]][[深度学习]]库，是广泛支持各种深度学习算法的[[运算]]框架<ref>{{cite web|url=http://venturebeat.com/2015/11/14/deep-learning-frameworks/|title=Want an open-source deep learning framework? Take your pick|last=Novet|first=Jordan|work=[[VentureBeat]]|date=2015-11-14|accessdate=2015-11-24|archive-date=2015-11-27|archive-url=https://web.archive.org/web/20151127042342/http://venturebeat.com/2015/11/14/deep-learning-frameworks/|dead-url=no}}</ref>。Deeplearning4j可以实施的技术包括[[受限玻尔兹曼机]]、[[深度置信网络]]、深度自动编码器、堆叠式降噪自动编码器、[[循环神经张量网络]]，以及[[word2vec]]、doc2vec和GloVe。这些算法全部包括[[分布式]][[并行]]版本，与[[Hadoop]]和[[Apache Spark|Spark]]集成。<ref>{{cite web|url=https://www.youtube.com/watch?v=LCsc1hFuNac&feature=youtu.be|title=Adam Gibson, DeepLearning4j on Spark and Data Science on JVM with nd4j, SF Spark @Galvanize 20150212|last=TV|first=Functional|work=SF Spark Meetup|date=2015-02-12|accessdate=2015-03-01|archive-date=2015-12-26|archive-url=https://web.archive.org/web/20151226235648/https://www.youtube.com/watch?v=LCsc1hFuNac&feature=youtu.be|dead-url=no}}</ref>[[Skymind]]是Deeplearning4j的商业支持机构。

== 简介 ==
Deeplearning4j基于广泛使用的编程语言Java——同时也兼容[[Clojure]]，并且包括[[Scala]]的API。它由自有的开源数值计算库[[ND4J]]驱动，可使用CPU或GPU运行。<ref name="om">{{cite web|first=Derrick|last=Harris|title=A startup called Skymind launches, pushing open source deep learning|work=[[GigaOM.com]]|date=2014-06-02|url=http://gigaom.com/2014/06/02/a-startup-called-skymind-launches-pushing-open-source-deep-learning/|accessdate=2014-06-29|archive-date=2014-06-28|archive-url=https://web.archive.org/web/20140628031148/http://gigaom.com/2014/06/02/a-startup-called-skymind-launches-pushing-open-source-deep-learning/|dead-url=no}}</ref><ref name="vb">{{cite web|first=Jordan|last=Novet|title=Skymind launches with open-source, plug-and-play deep learning features for your app|date=2014-06-02|url=http://venturebeat.com/2014/06/02/skymind-launches-with-open-source-plug-and-play-deep-learning-features-for-your-app//|accessdate=2014-06-29|archive-date=2014-06-28|archive-url=https://web.archive.org/web/20140628065946/http://venturebeat.com/2014/06/02/skymind-launches-with-open-source-plug-and-play-deep-learning-features-for-your-app/|dead-url=no}}</ref> Deeplearning4j是[[开源]]项目<ref>{{cite web|title=Github Repository|url=https://github.com/agibsonccc/java-deeplearning|accessdate=2016-03-27|archive-date=2019-09-24|archive-url=https://web.archive.org/web/20190924233626/https://github.com/eclipse/deeplearning4j|dead-url=no}}</ref>，主要由位于旧金山的一支[[机器学习]]团队开发，团队由Adam Gibson领导。<ref name="deeplearning4j.org">{{cite web|url=http://deeplearning4j.org/|title=deeplearning4j.org|accessdate=2016-03-27|archive-date=2016-03-30|archive-url=https://web.archive.org/web/20160330205423/http://deeplearning4j.org/|dead-url=no}}</ref><ref>{{cite web|title=Crunchbase Profile|url=http://www.crunchbase.com/person/adam-gibson|accessdate=2016-03-27|archive-date=2017-07-31|archive-url=https://web.archive.org/web/20170731113033/https://www.crunchbase.com/person/adam-gibson|dead-url=no}}</ref>Deeplearning4j是谷歌Word2vec页面上列出的唯一一个在Java环境下实施Word2vec的开源项目。<ref>{{cite web|title=Google Code|url=https://code.google.com/p/word2vec/|accessdate=2016-03-27|archive-date=2016-03-10|archive-url=https://web.archive.org/web/20160310194120/https://code.google.com/p/word2vec/|dead-url=no}}</ref>

Deeplearning4j已经用于多项商业和科研应用。其代码由[[GitHub]]<ref>{{cite web|url=https://github.com/piskvorky/gensim|title=GitHub - piskvorky/gensim: Topic Modelling for Humans|author=piskvorky|date=|work=GitHub|accessdate=2016-03-27|archive-date=2016-02-01|archive-url=https://web.archive.org/web/20160201030941/https://github.com/piskvorky/gensim|dead-url=no}}</ref>托管，并在[[Google groups|谷歌小组]]<ref>{{cite web|url=https://groups.google.com/forum/#!forum/deeplearning4j|title=Google Groups|author=|date=|work=google.com|accessdate=2016-03-27|archive-date=2011-01-22|archive-url=http://arquivo.pt/wayback/20110122130054/https://groups.google.com/forum/#!forum/deeplearning4j|dead-url=no}}</ref>上设有支持论坛。

这一框架是可组合的，即受限玻尔兹曼机、卷积网络、自动编码器、递归网络等浅层神经网络可以相互叠加，组合成不同类型的深度网络。

== 分布式 ==
Deeplearning4j的定型以集群进行。神经网络通过迭代化简平行定型，可以在Hadoop/YARN以及Spark上运行。<ref name="deeplearning4j.org"/><ref>{{cite web|url=https://github.com/emsixteeen/IterativeReduce|title=Iterative reduce|accessdate=2016-03-27|archive-date=2018-06-10|archive-url=https://web.archive.org/web/20180610234359/https://github.com/emsixteeen/IterativeReduce|dead-url=no}}</ref>Deeplearning4j还与Cuda内核集成，进行纯GPU操作，可使用分布式GPU运行。

== Java虚拟机中的科学计算 ==
Deeplearning4j包括使用ND4J的N维数组类，可在Java和Scala中进行科学计算，类似于[[Numpy]]为[[Python]]提供的功能。其基础是线性代数库，可有效支持生产环境中的矩阵操作。

== 用于机器学习的Canova向量化库 ==
Canova可将各类文件格式和数据类型向量化，所用的输入/输出格式系统近似于Hadoop的MapReduce。Canova目前仍在开发中，设计目标是实现CSV、图像、声音、文本和视频的向量化。Canova可以从命令行使用。
版本0.4.0之后，Canova库已合并到 DataVec当中。

== 文本与NLP ==
Deeplearning4j包括一个向量空间模型和主题模型工具包，在Java中实施，与并行GPU集成以提高表现。这是专门为处理大量文本而设计的。

Deeplearning4j可实施tf–idf、深度学习以及Mikolov的word2vec算法、doc2vec和GloVe－在Java中再实施并优化。它依靠t-SNE生成视觉化的文字云。

== 实际应用情景与集成 ==
Deeplearning4j的实际应用情景包括金融行业<ref>{{cite web|url=http://www.skymind.io/finance/|title=FINANCE & FRAUD|author=|date=|work=Skymind|access-date=2016-03-27|archive-url=https://web.archive.org/web/20160310082208/http://www.skymind.io/finance/|archive-date=2016-03-10|dead-url=yes}}</ref>的欺诈侦测、制造业等行业中的异常检测、电子商务与广告业的推荐系统、图像识别等。Deeplearning4j已与RapidMiner和Prediction.io等其他机器学习平台集成。

== 相关库 ==
* [[OpenNN]]，一个用C++语言编写的深度学习开源神经网络库。
* [[Torch_(机器学习框架)|Torch]]，一个用Lua语言编写的开源框架，广泛支持各类机器学习算法。
* [[Theano]]，一个为Python开发的开源深度学习库。
* [[Neuroph]]

== 参见 ==
* [[深度学习框架比较]]
* [[人工智能]]
* [[机器学习]]
* [[深度学习]]
* [[卷积神经网络]]

== 参考文献 ==
{{Reflist}}

== 外部链接 ==
* {{Official|http://www.deeplearning4j.org/zh-index}}
* {{Cite web|url=https://github.com/deeplearning4j|title=Deeplearning4j Github Repositories|access-date=2016-03-27|archive-date=2021-02-27|archive-url=https://web.archive.org/web/20210227094159/https://github.com/deeplearning4j|dead-url=no}}
* {{Cite web|url=http://deeplearning4j.org/compare-dl4j-torch7-pylearn.html|title=Deeplearning4j vs. Torch vs. Caffe vs. Theano|access-date=2016-03-27|archive-date=2016-05-30|archive-url=https://web.archive.org/web/20160530104456/http://deeplearning4j.org/compare-dl4j-torch7-pylearn.html|dead-url=yes}}<cite class="citation web" contenteditable="false"></cite><cite class="citation web" contenteditable="false">.</cite><span class="Z3988" title="ctx_ver=Z39.88-2004&rfr_id=info%3Asid%2Fen.wikipedia.org%3ADeeplearning4j&rft.btitle=Deeplearning4j+vs.+Torch+vs.+Caffe+vs.+Theano&rft.genre=unknown&rft_id=http%3A%2F%2Fdeeplearning4j.org%2Fcompare-dl4j-torch7-pylearn.html&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Abook" contenteditable="false"> </span>
* {{Cite web|url=http://deeplearning4j.org/canova.html|title=Canova: A General Vectorization Lib for Machine Learning|access-date=2016-03-27|archive-date=2016-06-12|archive-url=https://web.archive.org/web/20160612171403/http://deeplearning4j.org/canova.html|dead-url=yes}}
* {{Cite web|url=http://nd4j.org|title=ND4J: N-Dimensional Arrays for Java and Scala, A Linear Algebra Library|access-date=2022-03-11|archive-date=2020-06-01|archive-url=https://web.archive.org/web/20200601013308/https://nd4j.org/|dead-url=yes}}
* {{Cite web|url=https://flink.apache.org/|title=Apache Flink|access-date=2016-03-27|archive-date=2018-12-26|archive-url=https://web.archive.org/web/20181226025402/https://flink.apache.org/|dead-url=no}}
* {{Cite web|url=http://www.oraclejavamagazine-digital.com/javamagazine/may_june_2015?sub_id=DJ9kzXBnuXELe#pg58|title=Java Magazine by Oracle: Deep Learning in Java|access-date=2016-03-27|archive-date=2015-09-24|archive-url=https://web.archive.org/web/20150924061240/http://www.oraclejavamagazine-digital.com/javamagazine/may_june_2015?sub_id=DJ9kzXBnuXELe#pg58|dead-url=no}}
* {{Cite web|url=https://gitter.im/deeplearning4j/deeplearning4j|title=Deeplearning4j Community of Gitter Chat|access-date=2016-03-27|archive-date=2020-08-03|archive-url=https://web.archive.org/web/20200803164820/https://gitter.im/deeplearning4j/deeplearning4j|dead-url=no}}

{{深度学习软件}}

[[Category:人工神经网络]]
[[Category:叢集計算]]
[[Category:Scala平台軟體]]
[[Category:用Java編程的自由軟體]]
[[Category:Hadoop]]
[[Category:图像处理]]
[[Category:美國資訊科技公司]]
[[Category:JVM程式語言]]
[[Category:Java函式庫]]
[[Category:Java平台]]
[[Category:机器学习]]
[[Category:自然語言處理]]
[[Category:使用Apache许可证的软件]]
[[Category:开源人工智能]]

Deeplearning4j - 版本历史

imported>JimGrassroot：​使用HotCat已添加Category:开源人工智能

imported>JimGrassroot：使用HotCat已添加Category:开源人工智能