Overview of Popular Deep Learning Libraries Across Programming Languages

Python : Theano enables array‑based mathematical expression definition and computation, forming the basis for many libraries such as Keras (modular neural‑network library with GPU/CPU support), Pylearn2 (models and training algorithms), Lasagne (lightweight network wrapper), Blocks, and others including Caffe (C++ core with Python API), nolearn, Gensim, Chainer, deepnet, Hebel, CXXNET, DeepPy, DeepLearning, and Neon.

Matlab : Includes ConvNet (convolutional neural networks), DeepLearnToolBox (DBN, stacked auto‑encoders, CNN), cuda‑convet (C++/CUDA CNN code), and MatConvNet (vision‑focused CNN toolbox).

C++/CPP : eblearn (open‑source C++ machine‑learning library), SINGA (Apache‑backed distributed training), NVIDIA DIGITS (browser‑based deep‑network development and visualization), and Intel® Deep Learning Framework (accelerated CNNs on Intel platforms).

Java : ND4J (scientific computing for JVM), Deeplearning4j (commercial‑grade distributed deep learning), and Encog (machine‑learning framework covering SVMs, neural nets, genetic algorithms).

JavaScript : Convnet.js runs entirely in the browser, enabling training of deep learning models without external dependencies.

Lua : Torch provides a flexible scientific computing framework with LuaJIT scripting and C/CUDA back‑end.

Julia : Mocha offers a Caffe‑inspired deep learning framework with modular structure and efficient SGD implementation.

Lisp : Lush (Lisp Universal Shell) supplies object‑oriented programming and extensive deep learning libraries.

Haskell : DNNGraph is a domain‑specific language for generating deep neural network models.

.NET : Accord.NET is a comprehensive machine‑learning framework written in C# covering vision, audio, and signal processing.

R : The darch package enables construction of deep neural networks with pre‑training and fine‑tuning, while deepnet implements various deep learning algorithms such as RBM, DBN, and auto‑encoders.

Original source: CSDN Big Data.