Excel's macro/VBA language, with its 277 commands, including if-then-else programming commands, is as powerful as BASIC. It allows developers and power users to create sophisticated custom applications with added neural-net intelligence.
Excel's built-in Q+E and Solver also allow for direct access to a variety of databases (such as dBASE III and IV and Oracle), with fast selection of training subsets, goal-seeking computations to fine-tune performance, and sensitivity analysis to determine the most relevant variables, using Excel's interactive 3D graphics to display results.
For example, to test different investment scenarios, you could click on a bar graph and drag a column up or down to see the effect of changing oil prices on a stock. Excel would dynamically change related values, based on the current formula (using its goal seeking Solver program where multiple variables are involved).
Since these variables are hotlinked to Braincel, the neural net output would also change, displaying the new forecast or advice. This output can then be passed in real time to Excel - and other programs that support DDE - for hotlinked modeling and analysis. This kind of real-time, seamless, neural-net-aided computation provides a new level of interactive desktop analytic power to decision-makers.
Excel includes a macro recorder that further speeds application development by recording keystrokes and mouse clicks, thus shortening development time. Braincel's own macro library interfaces seamlessly into the spreadsheet.
Braincel is also available in a royalty-free runtime version that allows powerful and seamlessly-integrated applications to be delivered to end users.
Braincel includes an optional professional mode that allows for full control of parameters that fine-tune the neural net. These include number of hidden layers and nodes, learning rate, error rate, training cycles, connection weights, randomizing of weights, batching of weight adjustments, and momentum. The neural net's output can be displayed as text or as scaled (0 to 1) or unscaled (real-world) numbers.
Braincel includes extensive monitoring capabilities. For example, an icon shows average error and training time, color-coded for status (black when open, red, blue or green when training; when training ends, it turns black). A "notes" box lets you attach text to your neural net file, so you can keep track of pertinent information, such as what the neural net was trained to do. Training can be stopped at any time and the current net "snapshot" stored, allowing training to continue at a future time.
Several charts are included to monitor Braincel's learning process. Several utilities also help judge the accuracy of the net.
Braincel Questions and Answers