Deep Learning Library finishing in various programming languages

Mark, let's study for a moment.

Original address: http://www.csdn.net/article/2015-09-15/2825714

Python

1. Theano is a Python class library that uses array vectors to define and calculate mathematical expressions. It makes it easy to write deep learning algorithms in a python environment. On top of it, many classes of libraries have been built.

1.Keras is a compact, highly modular neural network library that is designed to reference torch, written in Python, to support the invocation of GPU and CPU-optimized Theano operations.

2.PYLEARN2 is a library that integrates a large number of deep learning common models and training algorithms, such as random gradient descent. Its library of functions is based on Theano.

3.Lasagne is a lightweight package library for building and training neural networks, based on Theano. It follows the principles of brevity, transparency, modularity, practicality and specificity.

4.Blocks is also a framework based on Theano to help build neural networks.

2. Caffe is the framework of deep learning, which focuses on the expression of code, the speed of operation and the degree of modularity. It was developed jointly by the Berkeley Vision and Learning Center (Berkeley Vision and Learning Center, BVLC) and community members. Google's Deepdream project is done based on the Caffe framework. This framework is a C + + library that uses the BSD license and provides a Python invocation interface.

3. Nolearn includes a large number of packages and abstraction interfaces for existing neural network libraries, the famous lasagne, and some common modules for machine learning.

4. Genism is also a deep learning gadget written in Python that uses efficient algorithms to process large-scale text data.

5. Chainer a bridge between the theoretical algorithm and the practical application of deep learning. It is characterized by strong, flexible, intuitive, and is considered a flexible framework for deep learning.

6. Deepnet is a library of deep learning algorithms based on the GPU, developed using Python language to implement Feedforward neural Networks (FNN), restricted Boltzmann machines (RBM), depth belief network (DBN), self-encoder (AE), Algorithms such as deep Boltzmann Machines (DBM) and convolutional neural Networks (CNN).

7. Hebel is also a python library of deep learning and neural networks, which controls the support of Cuda GPU acceleration through Pycuda. It implements the most important types of neural network models, and provides a variety of activation functions and model training methods, such as momentum, Nesterov momentum, dropout, and early stopping methods.

8. Cxxnet is a fast and concise distributed deep learning framework based on Mshadow development. It is a lightweight, extensible C++/cuda Neural Network Toolkit that provides a friendly Python/matlab interface for training and prediction.

9. Deeppy is a deep learning framework based on NumPy.

Deeplearning is a library of deep learning functions developed in combination with C + + and Python.

Neon is a deep learning framework for the Nervana System, developed using Python.

Matlab

1. Convnet convolutional Neural Network is a kind of deep learning classification algorithm, which can learn the useful characteristics from the original data and realize it by adjusting the weight value.

2. Deeplearntoolbox is a Matlab/octave toolkit for deep learning that includes algorithms such as depth Belief network (DBN), stacked self-encoder (stacked AE), convolutional Neural Network (CNN), and more.

3. Cuda-convet is a convolutional neural network (CNN) code that is also suitable for feedforward neural networks and is used for c++/cuda operations. It can model multilayer neural networks of any depth. As long as there is a network structure to the non-circular graph can. The training process uses the inverse propagation algorithm (BP algorithm).

4. Matconvnet is a convolutional neural network (CNN) MATLAB Toolbox for computer vision applications. It is simple and efficient, capable of running and learning the most advanced machine learning algorithms.

CPP

1. Eblearn is an open source machine learning C + + packaging library developed by Yann LeCun, a New York University machine Learning Laboratory. It uses an energy-based model to implement convolutional neural networks and provides visual interface (GUI), examples, and demonstration tutorials.

2. Singa is a project supported by the Apache Software Foundation, which is designed to provide a common distributed model training algorithm on existing systems.

3. NVIDIA digits is a new set of systems for developing, training and visualizing deep neural networks. It presents the powerful features of deep learning in a browser interface, enabling data scientists and researchers to visualize neural network behavior in real time and quickly design a deep neural network that is best suited to the data.

4. The Intel®deep Learning Framework provides a unified platform for Intel® platform accelerated deep convolutional neural networks.

Java

1. N-dimensional Arrays for Java (nd4j) is a library of scientific computing functions for the JVM platform. It is mainly used in the product, that is, the function of the design needs of the operation speed, storage space is the most province.

2. Deeplearning4j is the first commercial-grade open-source distributed deep Learning Library, written in Java and Scala. It is designed to be used in a commercial environment, rather than as a research tool.

3. ENCOG is a high-level machine learning framework that includes support vector machines, artificial neural networks, genetic programming, Bayesian networks, and Hidden Markov models, as well as genetic algorithms.

JavaScript

1. Convnet.js, written by JavaScript, is a complete library of training deep learning models (mainly neural networks) completed in the browser. No need for other software, no compilers, no installation packages, no GPU, no need for even effortless.

Lua

1. Torch is a scientific computing framework that is widely applicable to a variety of machine learning algorithms. It is easy to use, developed with Fast scripting language Luajit, and the underlying is C/cuda implementation. Torch is based on the LUA programming language.

Julia

1. Mocha is the deep learning framework of Julia, inspired by the C + + framework Caffe. The efficient implementation of the general stochastic gradient Solver and the Universal module in the Mocha can be used to train the depth/shallow layer (convolutional) neural network, which can be completed by the (stacked) self-encoder with unsupervised pre-training (optional). Its advantages include a modular structure, an upper interface, and possibly more features such as speed, compatibility, and more.

Lisp

1. Lush (Lisp Universal Shell) is an object-oriented programming language for a broad range of researchers, experimenters, and engineers interested in large-scale numerical and graphical applications. It has a library of machine learning functions, which contains a rich library of deep learning.

Haskell

1. Dnngraph is a domain-specific language (DSL) used by Haskell for the generation of deep neural network models.

. NET

1. Accord.net is written entirely in C #. NET machine learning framework, including the class library for audio and image processing. It is the complete framework of the product-level for computer vision, computer audio, signal processing, and statistical applications.

R

1. The Darch package can be used to generate a multilayer neural network (depth structure). Training methods include pre-training for contrast divergence and fine tuning of well-known training algorithms such as reverse propagation or conjugate gradient methods.

2. Deepnet implements many deep learning frameworks and neural network algorithms, including reverse propagation (BP), restricted Boltzmann machine (RBM), depth belief network (DBP), depth self-encoder (deep Autoencoder), and more.

Deep Learning Library finishing in various programming languages