OpenANN: Comparison with Other Libraries

OpenANN 1.1.0

An open source library for artificial neural networks.

All Classes Namespaces Files Functions Variables Typedefs Enumerations Enumerator Friends Macros Pages

	Website	Languages	License	Focus
OpenANN	Link	C++, Python	GPL	An easy to use, extend and understand C++ library for artificial neural networks with Python bindings. It is designed to provide efficient algorithms for small neural networks that solve standard machine learning problems as well as for deep learning algorithms. It has only a small number of dependencies and thus, it is suited to run on limited systems.
FANN	Link	C, C++, Python, Ruby, C#, Java, Perl, Delphi, ... (almost everything)	LGPL	A well-known and widly used library for feedforward neural networks written in C with bindings for almost any language. FANN has limited functionality but is easy to use and robust. It usually used for small datasets and applications.
pylearn2	Link	Python	BSD	State of the art implementations of many deep learning algorithms that are build upon Theano in Python.
PyBrain	Link	Python	BSD	A huge python library that contains many types of neural networks for supervised, unsupervised and reinforcement learning, e.g. recurrent neural networks (long short-term memory, reservoir), restricted boltzman machines, self-organizing maps, ...
cuda-convnet	Link	C++/CUDA, Python	BSD	State of the art CUDA library for large CNNs. It is really fast, has python bindings and has been used to achieve state of the art results on the ImageNet dataset. Many types of layers and activations functions are implemented and new types can be added easily. However, using it and can be very difficult.
EBLearn	Link	C++	BSD	A state of the art library for energy based models and e.g. convolutional neural networks with a large infrastructure. It allows to use CUDA code.
Torch7	Link	Lua	BSD	Torch 7 is a very efficient tensor library that supports speed-up with SSE, OpenMP, CUDA and is designed to implement machine learning algorithms. It supports automatic differentiation and is very similar to Theano. It has many packages, e.g. for deep learning models and optimization algorithms.