October 2000 Newsletter

I. A referral $$$ program for you

Do you have friends or colleagues who could benefit from one of our major products (any version of NeuroShell, GeneHunter, or the Trilogy)? Refer that person to us and get yourself a $249 credit towards the purchase of any of our products, addons, or upgrades! Yes, if they purchase any of our major products between now and the first of the year and keep it, you’ll save $249 on your next purchase! Just make sure you or the person referred informs us about the referral within a week of purchase. Trader users please note that the $249 will completely pay for one of our Trader addon products!


II. The AI Trilogy(Tm) is shipping!

Attention consultants: here is your chance to make lots of money!

Our September newsletter announced the AI Trilogy, a new packaging of our flagship business and scientific artificial intelligence products (NeuroShell Predictor, Classifier, and GeneHunter). The AI Trilogy is now shipping.

You will recall that the AI Trilogy was meant to solve a particular type of problem we called the AI Trilogy Problem:

Your company is formulating some mixture: maybe a drug, a bonding compound, an animal feed, foodstuff or beverage, plastic or other packaging material, cosmetic or perfume, lubricating compound, cleaning compound or solvent, paint, insecticide or herbicide, a manufactured item or process, or even problems such as marketing strategies. Your boss or your client has given you the job of finding the optimum mixture of components or process variables, the combination of components and processing methods that produces the best product (components are the percentages of each of the elements of the mixture and the process variables are factors in the building process such temperature, curing time, etc.)

The way you can solve the AI Trilogy problem is documented in the AI Trilogy help files, along with two examples.

The AI Trilogy costs you only $995, plus shipping. You can charge your clients 20 to 100 times that amount for every AI Trilogy problem you solve.

Please see www.wardsystems.com for more details, or call our sales department for a personal explanation of the AI Trilogy problem.


III. Asset allocation

Many of our users have asked us how to do asset allocation or portfolio management. The usual goal is to maximize profit while minimizing risk. That problem is generally not one that the NeuroShell series was built to tackle. However, such a problem is perfect for GeneHunter, our genetic algorithm optimizer.

GeneHunter contains an example of asset allocation wherein high returning assets are allocated to minimize risk. There are many other similar methods published in which you can minimize risk and maximize profit. You’ve probably seen them in books and articles. What they have in common is that they require optimization (usually in a spreadsheet) of some complex formula(s) in which risk in minimized while profit is maximized. That is where you can use GeneHunter.

Traditional optimizers break down with nasty problems (when the data is not continuous and not differentiable) whereas GeneHunter can optimize almost anything.

By the way, GeneHunter examples also include examples of portfolio allocation, building optimal rules and optimal indicators for market predictions, clustering, building a neural net in Excel, and building non-linear mathematical approximating functions.

IV. It was the best of times, it was the worst of times…

The year 2000 before March was certainly the best of times for NASDAQ investors. After March, it was the worst of times, unless you had the NeuroShell Trader. Volatile markets are actually the best of times to make profitable models. Why? Because when prices are thrashing up and down and up and down again, there are lots of opportunities for neural nets to pick up cyclic moves. After all, lately there arguably have been only two major variables driving the market: fear and greed! After greed drives high tech prices up, fear drives them back down again. And if you are fortunate enough to own the DayTrader, you can use hourly or even smaller bars so you don’t have to train back prior to March. We love these times!


V. Trader Tips. Did you know…

1. Did you know that the NeuroShell Trader Professional and DayTrader Professional contain a technology that engineers have used for decades to preprocess noisy time series data? No, not wavelets, although we have them too. It is the Fast Fourier Transform (FFT). Fourier analysis is long proven technology, not “pie in the sky.” The Fourier Transform is based on the fact that it is possible to take any periodic time series and break it down into equivalent frequency components. We know that many users are searching the internet for the holy grail of time series analysis, but you might want to start with something well known to be effective. Example 2 of the Adaptive Turboprop 2 Add-on is an illustration of one way to use the FFT.

2. Did you know that the NeuroShell Trader has the capability of automatically and without intervention building (and optimizing in the Pro) models for many different issues? All you have to do is define the model or optimization once for one issue and the Trader will then do the same thing for a whole series of issues while you eat, sleep, or go fishing.

The way this is done is to build your chart as usual, using indicators, predictions, and trading strategies. Then, using the menu item Add/Remove Chart Pages from the Edit menu, you select more issues for which you want to build the same type of model. Once they have been selected, the Trader will ask you if you want to apply your model to the newly chosen issues. Say yes and you are on your way.

For more information, log onto www.ward.net and read the tip “NeuroShell Trader Batch Processing Capability.”

3. Did you know that you can continue optimization even after optimization stops? Most of you probably knew that you can stop optimization early in the Professional versions, but you can also make it go longer. The genetic algorithm optimizer makes an intelligent decision about when to stop optimizing, because there is no way to ever know for sure when enough is enough. And although we sometimes recommend stopping early if you feel you are “overfitting”, there could also be occasions when the out-of-sample doesn’t do well because optimization didn’t run long enough to find the real relationships. (No, there is no science here, and no way to know in advance whether (1) you went too long, (2) not long enough, or (3) there are just no relationships to be found!)

To continue when optimizing in the Prediction Wizard, just hit the retrain button. If you are optimizing in the Trading Strategy Wizard, just hit the Rebacktest button. In both cases, you’ll then get a message asking if you want to start optimizing using the “last optimal settings.” If you say yes, the new optimization run will be “seeded” with the best answer already found on the last run, so that you aren’t really starting fresh. You can do that as many times as you wish.


Was this article helpful?

Related Articles