What is MemBrain?

MemBrain is a powerful graphical neural network editor and simulator for Microsoft Windows, supporting artificial neural networks of arbitrary size and architecture.

Some core features of MemBrain are:

• Powerful, easy-to-learn and intuitive graphical editor and simulator for Artificial Neural Networks (ANN)
• Transfer of trained neural nets into production systems by dll or automatically generated C or C++ Code
• 'On-The-Fly'-Validation during training using separate validation data
• Supervised and unsupervised learning
• Powerful neuron and link model, time invariant or dynamic through parametrization. Supports everything from compact function approximators up to complex convolutional neural nets with pooling layers.
• Integrated high performance object oriented scripting language
• Integrated optional 256 Bit AES encryption to help protect your IP and Know How

What is membrain used for?

MemBrain originally had been developed to provide maximum flexibility for development and study of artificial neural networks combined with an intuitive graphical user interface. It is currently used by many universities for research and teachings, however, during the last years its use in industrial manufacturing and technical control applications has stepped up continuously.

The majority of applications hereby is considered with the challenge to map correlations between inputs and outputs of unknown or not sufficiently known systems into artificial neural networks.

The goal of this is to be able to predict the outputs of the system under test on basis of its inputs  without having to know the corresponding transfer functions. Knowledge of these functions is often not achievable at all or at least not accessible with justifiable efforts.
Solely required to achieve this goal is a sufficient amount of real world input and output data pairs which might be captured from the real system or might already be available from historic records. These data can be imported into MemBrain and be used to train the corresponding artificial neural network.  

Once the neural net has been designed and trained successfully it can be used to approximate the real system for the use in forecasts or other calculations: New values are applied to the inputs of the net -  the target output values are calculated by the net and accessible through its output ports.

Using MemBrain, such a net can be easily transferred into production systems, no matter if these are PC based or embedded systems. PC based systems can use the dll version of MemBrain to load the trained nets from files and make use of them in an efficient way. The dll also allows to continue the training of the nets within the user's application.
For embedded systems MemBrain brings an integrated C and C++ source code generator to generate code that is aligned with the typical requirements of embedded systems: Small memory footprint and avoidance of dynamic memory allocation.

What does it cost?

MemBrain is freeware. It can be used free of charge for private as well as for commercial purposes.