What's being said about Braincel

"Sniffs out relationships that can elude conventional analysis"
Business Week
"Solves problems that used to be attacked by expert systems or statistical analysis"
Wall Street Journal
"The first neural network implementation the enduser can pick up and learn how to use very quickly"
PC Week
Barron's dated December 14, 1992 had an article by Avener Mandelman entitled "A Money Manager's Love Affair With Neural Network Programs." In the article he described Braincel at $249.00 as "A favorite of mine"
"I use Braincel to forecast stocks, industry groups, Fidelity Funds, and the market as a whole. Braincel is easy to use, fast and has superior ability to recognize non linear patterns in my data" Philip Erlanger, editor of Next Generation Investor .
"I am a Fee Based Portfolio Manager investing primarily in no load mutual funds. During the last 7 years I have worked to refine the asset allocation models, constantly changing the algorithms I use. Three or four years ago I started experimenting with Braincel believing that by using this advanced tool I would improve my end results. I track and use over 200 indices and review 1800+ mutual funds, searching for significant correlation's over different time periods. In the beginning I had very mixed results with Braincel until I started preprocessing my data... Using Braincel I preprocess mutual fund returns and prices to eliminate any anomalies resulting from non-economic influences. This gives me a range of normalized real returns that I use as the basis for forecasting. Forecast indices come from a variety of economists, information services, and my own estimates...This gives very good indications of buy/sell signals for strategic investing, if the price is above the forecast by a predetermined % then sell and vice versa to buy.
For the dynamic asset allocation models I built a 20 year historic quarterly model by hand based on the above normalized data, training Braincel with this data I then forecast the models allocation for the next quarter and reallocate all my portfolios accordingly.
I invest based on longterm financial scenarios and Braincel shines in this area. The returns to my clients have consistently been well above the appropriate market averages and continue to bring me new clients. I look forward to continuing improvement in your product (Braincel) and hence in mine." Steven Sunstein, United Pacific Securities, Inc.
"As chemical engineers we are users of the technology not neural network scientists. Braincel gives us a tool that works for us without a lot of overhead learning the package. We have yet to stump it even with some huge data sets. Other packages targeted for chemical engineering applications cost as much as $50,000.

Much of our work requires preprocessing the data to form inputs that make sense from an chemical engineering standpoint. By staying in the Excel environment, Braincel simplifies this part of the work unlike some of the other packages which require changing back and forth between environments.

The models we develop of chemical processes provide our customers with " Virtual Analyzers" costing 5-10% of a field proven hardware process analyzer. And, frequently a field type proven device is not even available."

The second generation of our product, Virtual Analyzer, is now packaged with Braincel as the modeling engine. The first generation required our staff to build the models and be on site to migrate them to the process control system. Obviously, being able to "shrink wrap" a complete package for the end user would be a big benefit for all concerned. While we had chosen Braincel to develop Virtual Analyzer, there was a serious question of our ability to write instructions which we could reasonably expect our customers to use in field locations. We were looking at satisfying developers possibly on their first application, having to work with Braincel, Virtual Analyzer and the control system vendors product. As we all know, in industry "they expect results".

Fortunately, the same characteristics of Braincel that lead us to select it for our first generation of Virtual Analyzer convinced us it was the neural network modeling program to package in our second generation. We could be reasonably confident our customers could develop the process models with Braincel. Even possibly complex data preprocessing would be expedited by working in the Excel environment already familiar to so many developers. Braincel's excellent documentation and tutorials reassured us that our customers would be able to acquire the modeling skills smoothly. Braincel's Excel interface meant that we could prepare instructions for the user to extract the model and migrate it to the control system via a series of templates and familiar edit commands.

Roland Hinkle
Pacific Technologies, LLC
Secrest Research provides technical services to semiconductor companies. These services assist the clients with improving product and manufacturing performance. The assistance includes the application of neural nets (Braincel in particular) to predict outcomes of manufacturing steps. Jerry Secrest tells us of particular applications:
  • We predict outcomes of manufacturing steps. For example we predict transistor parameters from in fab process data.
  • We predict final product characteristics - such as speed and operating voltage range of integrated circuits.
  • We screen out low performing product early in the process to save money
  • We predict outcome of complex processes to feedback to input variables. These complex processes include 'deposition of thin films via chemical vapor', plasm etching of geometries, etc.

We have also used Braincel to assist with diagnostics but found this to be not very satisfactory as most clients want to pin the problem on one variable. Neural nets are not very descriptive about isolating input variables and most problems that need neural nets are not one variable problems.
In the future we want to expand our services to:

  • Remove maverick parts
  • create large models predicting outcomes of many products and manufacturing variables
  • Check for low level shifts in manufacturing
  • Compare two supposedly identical manufacturing lines

Back to Home