The Model Manager includes:

• optimisation in a bayesian framework, the hyperparameters are also automatically adjusted.
• automatic relevance determination
• automatic construction of an optimum committee of models
• possibility to perform all the training as a background task, implying that others can use your workstation during the training, and even log you out.
• possibility to distribute the training on a network of machines to reduce computation time.
• ability to handle very large database. There is no size limit hard-wired in the program (but there is in your hardware). To date, the largest database used by one of our customers was of 100,000 lines with 30 input variables.
• possibility to export your models in a variety of formats:
  • self-contained DOS/Windows binaries
  • read-to-install packages for unix/linux
  • packages for the Predictor, the ideal tool to make predictions with existing models.

and nevertheless remains so simple to use that you could have results in 5 minutes without even knowing what is a neural network.