This article has multiple issues. Please help improve it or discuss these issues on the talk page . (Learn how and when to remove these template messages)
|
Original author(s) | Alex D. Black, Adam Gibson, Vyacheslav Kokorin, Josh Patterson |
---|---|
Developer(s) | Kondiut K. K. and contributors |
Preview release | 1.0.0-beta7 / 13 May 2020 [1] |
Repository | |
Written in | Java, CUDA, C, C++, |
Operating system | Linux, macOS, Windows, Android, iOS |
Platform | CUDA, x86, ARM, PowerPC |
Available in | English |
Type | Natural language processing, deep learning, machine vision, artificial intelligence |
License | Apache License 2.0 |
Website | www |
Part of a series on |
Machine learning and data mining |
---|
Eclipse Deeplearning4j is a programming library written in Java for the Java virtual machine (JVM). [2] [3] It is a framework with wide support for deep learning algorithms. [4] Deeplearning4j includes implementations of the restricted Boltzmann machine, deep belief net, deep autoencoder, stacked denoising autoencoder and recursive neural tensor network, word2vec, doc2vec, and GloVe. These algorithms all include distributed parallel versions that integrate with Apache Hadoop and Spark. [5]
Deeplearning4j is open-source software released under Apache License 2.0, [6] developed mainly by a machine learning group headquartered in San Francisco. [7] It is supported commercially by the startup Skymind, which bundles DL4J, TensorFlow, Keras and other deep learning libraries in an enterprise distribution called the Skymind Intelligence Layer. [8] Deeplearning4j was contributed to the Eclipse Foundation in October 2017. [9] [10]
Deeplearning4j relies on the widely used programming language Java, though it is compatible with Clojure and includes a Scala application programming interface (API). It is powered by its own open-source numerical computing library, ND4J, and works with both central processing units (CPUs) and graphics processing units (GPUs). [11] [12]
Deeplearning4j has been used in several commercial and academic applications. The code is hosted on GitHub. [13] A support forum is maintained on Gitter. [14]
The framework is composable, meaning shallow neural nets such as restricted Boltzmann machines, convolutional nets, autoencoders, and recurrent nets can be added to one another to create deep nets of varying types. It also has extensive visualization tools, [15] and a computation graph. [16]
Training with Deeplearning4j occurs in a cluster. Neural nets are trained in parallel via iterative reduce, which works on Hadoop-YARN and on Spark. [7] [17] Deeplearning4j also integrates with CUDA kernels to conduct pure GPU operations, and works with distributed GPUs.
Deeplearning4j includes an n-dimensional array class using ND4J that allows scientific computing in Java and Scala, similar to the functions that NumPy provides to Python. It's effectively based on a library for linear algebra and matrix manipulation in a production environment.
DataVec vectorizes various file formats and data types using an input/output format system similar to Hadoop's use of MapReduce; that is, it turns various data types into columns of scalars termed vectors. DataVec is designed to vectorize CSVs, images, sound, text, video, and time series. [18] [19]
Deeplearning4j includes a vector space modeling and topic modeling toolkit, implemented in Java and integrating with parallel GPUs for performance. It is designed to handle large text sets.
Deeplearning4j includes implementations of term frequency–inverse document frequency (tf–idf), deep learning, and Mikolov's word2vec algorithm, [20] doc2vec, and GloVe, reimplemented and optimized in Java. It relies on t-distributed stochastic neighbor embedding (t-SNE) for word-cloud visualizations.
Real-world use cases for Deeplearning4j include network intrusion detection and cybersecurity, fraud detection for the financial sector, [21] [22] anomaly detection in industries such as manufacturing, recommender systems in e-commerce and advertising, [23] and image recognition. [24] Deeplearning4j has integrated with other machine-learning platforms such as RapidMiner, Prediction.io, [25] and Weka. [26]
Deeplearning4j serves machine-learning models for inference in production using the free developer edition of SKIL, the Skymind Intelligence Layer. [27] [28] A model server serves the parametric machine-learning models that makes decisions about data. It is used for the inference stage of a machine-learning workflow, after data pipelines and model training. A model server is the tool that allows data science research to be deployed in a real-world production environment.
What a Web server is to the Internet, a model server is to AI. Where a Web server receives an HTTP request and returns data about a Web site, a model server receives data, and returns a decision or prediction about that data: e.g. sent an image, a model server might return a label for that image, identifying faces or animals in photographs.
The SKIL model server is able to import models from Python frameworks such as Tensorflow, Keras, Theano and CNTK, overcoming a major barrier in deploying deep learning models.
Deeplearning4j is as fast as Caffe for non-trivial image recognition tasks using multiple GPUs. [29] For programmers unfamiliar with HPC on the JVM, there are several parameters that must be adjusted to optimize neural network training time. These include setting the heap space, the garbage collection algorithm, employing off-heap memory and pre-saving data (pickling) for faster ETL. [30] Together, these optimizations can lead to a 10x acceleration in performance with Deeplearning4j.
Deeplearning4j can be used via multiple API languages including Java, Scala, Python, Clojure and Kotlin. Its Scala API is called ScalNet. [31] Keras serves as its Python API. [32] And its Clojure wrapper is known as DL4CLJ. [33] The core languages performing the large-scale mathematical operations necessary for deep learning are C, C++ and CUDA C.
Tensorflow, Keras and Deeplearning4j work together. Deeplearning4j can import models from Tensorflow and other Python frameworks if they have been created with Keras. [34]
Scala is a strong statically typed high-level general-purpose programming language that supports both object-oriented programming and functional programming. Designed to be concise, many of Scala's design decisions are intended to address criticisms of Java.
Clojure is a dynamic and functional dialect of the Lisp programming language on the Java platform.
Apache Mahout is a project of the Apache Software Foundation to produce free implementations of distributed or otherwise scalable machine learning algorithms focused primarily on linear algebra. In the past, many of the implementations use the Apache Hadoop platform, however today it is primarily focused on Apache Spark. Mahout also provides Java/Scala libraries for common math operations and primitive Java collections. Mahout is a work in progress; a number of algorithms have been implemented.
Eclipse Vert.x is a polyglot event-driven application framework that runs on the Java Virtual Machine.
Akka is a source-available toolkit and runtime simplifying the construction of concurrent and distributed applications on the JVM. Akka supports multiple programming models for concurrency, but it emphasizes actor-based concurrency, with inspiration drawn from Erlang.
Apache Spark is an open-source unified analytics engine for large-scale data processing. Spark provides an interface for programming clusters with implicit data parallelism and fault tolerance. Originally developed at the University of California, Berkeley's AMPLab, the Spark codebase was later donated to the Apache Software Foundation, which has maintained it since.
TensorFlow is a free and open-source software library for machine learning and artificial intelligence. It can be used across a range of tasks but has a particular focus on training and inference of deep neural networks.
The following table compares notable software frameworks, libraries and computer programs for deep learning.
Thought vector is a term popularized by Geoffrey Hinton, the prominent deep-learning researcher, which uses vectors based on natural language to improve its search results.
Keras is an open-source library that provides a Python interface for artificial neural networks. Keras acts as an interface for the TensorFlow library.
Apache MXNet is an open-source deep learning software framework that trains and deploys deep neural networks. It aims to be scalable, allows fast model training, and supports a flexible programming model and multiple programming languages. The MXNet library is portable and can scale to multiple GPUs and machines. It was co-developed by Carlos Guestrin at the University of Washington, along with GraphLab.
Caffe is a deep learning framework, originally developed at University of California, Berkeley. It is open source, under a BSD license. It is written in C++, with a Python interface.
In computer vision, SqueezeNet is the name of a deep neural network for image classification that was released in 2016. SqueezeNet was developed by researchers at DeepScale, University of California, Berkeley, and Stanford University. In designing SqueezeNet, the authors' goal was to create a smaller neural network with fewer parameters while achieving competitive accuracy.
PlaidML is a portable tensor compiler. Tensor compilers bridge the gap between the universal mathematical descriptions of deep learning operations, such as convolution, and the platform and chip-specific code needed to perform those operations with good performance. Internally, PlaidML makes use of the Tile eDSL to generate OpenCL, OpenGL, LLVM, or CUDA code. It enables deep learning on devices where the available computing hardware is either not well supported or the available software stack contains only proprietary components. For example, it does not require the usage of CUDA or cuDNN on Nvidia hardware, while achieving comparable performance.
LightGBM, short for light gradient-boosting machine, is a free and open-source distributed gradient-boosting framework for machine learning, originally developed by Microsoft. It is based on decision tree algorithms and used for ranking, classification and other machine learning tasks. The development focus is on performance and scalability.
CatBoost is an open-source software library developed by Yandex. It provides a gradient boosting framework which among other features attempts to solve for Categorical features using a permutation driven alternative compared to the classical algorithm. It works on Linux, Windows, macOS, and is available in Python, R, and models built using catboost can be used for predictions in C++, Java, C#, Rust, Core ML, ONNX, and PMML. The source code is licensed under Apache License and available on GitHub.
{{cite web}}
: CS1 maint: archived copy as title (link){{cite web}}
: CS1 maint: archived copy as title (link)