BrainMaker

What is BrainMaker ?

Designed by seven Caltech mathematicians and engineers, BrainMaker is for people who demand the most powerful and ease-of-use development tools. BrainMaker's carefully chosen patterns and automated functions save time and make design easier, even for computer beginners.

In addition to completely new features that provide tighter control over training, offer more data manipulations, offer an easy-to-use manual, show training progress, and allow you to remove neurons without destroying all connections, provide more statistics for analysis, and greater control over noise and output files.

Here are some details of brainmaker's new features:

1. Stop Training options are based on any number of training performance or test performance criteria you choose. Allows you to easily find a network that tests well and allows you to stop at a preset point.
2. Network Progress Monitor graphically shows how well the network is learning. It helps determine the level of accuracy you can expect, how long the training will be performed and how well it is at any given time.
3. You can train a neural network and then reduce or increase the number of hidden neurons without having to start training again. BrainMaker eliminates the least significant neuron. This helps to train a network that generalizes well.
4. Five new types of moving averages help you better see trends and smooth out noise in your data in the best way for you. Five new types of moving averages help you better see trends and smooth out noise in your data in the best way for you. Absolute values have been added to NetMaker for more manipulations.
5. More statistics are calculated and reported in the training and test statistics files. The RMS error and the average error show the average network performance during execution and whether any fact is too far outside the specified tolerance. Helps diagnose problems and determine the best performing network.
6. Different amounts of input noise can be specified for training, execution, or fact testing. Testing with more noise can provide an even better idea of generalization capability.
7. The completely revised manual with new tutorials lets you decide how much you want to learn and guide you through the trouble-free design process.
8. Different output files can be written for training, execution, and testing. Facilitates the control of results throughout the design process.
9. Set all unmarked NetMaker columns to input, pattern, or annotations. Facilitates the creation of BrainMaker files.

System Requirements

Integration with Microsoft Windows 95, including Windows Dynamic Data Exchange (DDE). With DDE, you can create neural network capability in your spreadsheet, database, graphical program. Windows multitasking lets you train a network in the background while you use your computer for something else.