Core AI Technology Details

A well designed and specialized neural network has the capability to model almost any process (assuming the process is not purely random). A key point is 'Well Designed' though. Some Neural Net (Neural Net) programs are doomed from the beginning caused by poor designs or the inability of the programmer to understand the relationships between inputs, outputs and intended use of the system. Here are some of the core details to ANNI that help make it one of the best Neural Net Investing programs available.

Prediction System

Artificial Intelligence:	3 Layered, Feed Forward Neural Net Genetic Optimization
Training Method:	Back Propagation with Gradient Descent Genetic Algorithms Optimize Parameters
Specialized Techniques:	Neurons Are Truly Separate From One Another With Each Training Individually Weight Decay Algorithm to Prevent Overtraining Momentum Term to Get Out of Local Minima Committee System of Multiple Neural Nets Proprietary, Highly Adaptive, Specialized Learning Algorithm. See 3D Graph
Purpose	For predictions of actual prices up to 90 days. Used to help determine movement of a stock. Works exceptionally well on non-volatile stocks such as blue chips that are not going through any major changes. Does not work as well on highly volatile stocks or ones going through recent turmoil because of new events taking place (so keep an eye on news to counter this). This is basically because the system is a pattern identifier and if patterns do not ever repeat themselves because of the volatile nature of a stock...it will hinder the systems capabilities of accurately predicting a move. This is standard with any pattern searching/identifying system. Luckily ANNI has tools built in to preprocess stocks to let you know which ones are good and which ones are not saving you the time of even training on ones that will not work will with ANNI. And it also has safety checks for current files in case the stocks characteristics change significantly enough to warrant your attention.

Moving Average Trade System

Artificial Intelligence:	Genetic Optimization
Training Method:	Genetic Algorithms Optimize Parameters
Specialized Techniques:	Partitioned data with subset training to prevent localized hot spots from distorting the systems convergence Use of a spread allowance to reduce false signals generated by small temporary crossovers.
Purpose	For generation of trade signals to assist in trading strategies. Works best on non-volatile stocks such as blue chips that are not going through any major changes. Does not work as well on highly volatile stocks or ones going through recent turmoil because of new events taking place. This is basically because the system uses Moving Averages for the computations which are inherently lagged indicators (although genetically optimized to reduce this effect). This means it will react slower to fast changes then one may want (keeping an eye on current news will counter this). Luckily ANNI has tools built in to Preprocess stocks to let you know which ones are good and which ones are not. And it also has safety checks for current files in case the stocks characteristics change significantly enough to warrant your attention.

Genetically Trained Neural Nets (Advanced) Trade System

Artificial Intelligence:	3 Layered, Feed Forward Neural Net Genetic Training
Training Method:	Genetically Trained Neural Net Weights
Specialized Techniques:	Partitioned data with subset training to prevent localized hot spots from distorting the systems convergence Nets are Genetically Trained towards a target goal rather then a target value allowing the AI system to truly come up with it's own customized, successful trade system rather then one modeled after a possible detrimental system with restrictions (since not all standard trade systems work for all stocks) Proprietary Neural Net structure that can only be trained using Genetic Training which considerably increases the capabilities of the Neural Net.
Purpose	For generation of trade signals to assist in trading strategies. Very powerful system that has virtually no limitations to the condition of the stock with two exceptions. Can train on securities that only have Closing values but is designed and will work best with securities that have Close, Open, High, Low and Volume. Otherwise, as you can imagine, less input data means less accuracy (ANNI will of course warn you if you create a file using the GN Trade System that doesn't have all available data inputs that it would like). As with all systems, cannot predict critical events that can have major impacts on a stock (i.e. such as the CEO stealing from the company or lying about the company's income). Simply keeping an eye on current news will counter this.

Real-Time Predictor System

Artificial Intelligence:	Coupled, 3 Layered, Feed Forward Static Neural Nets
Training Method:	None Required
Specialized Techniques:	Nets use a Proprietary structure that has been specifically engineered to work with Real-Time Streaming data via two separate Neural Nets. One Net processes quotes and the other processes ticks for maximum efficiency and precision. Outputs/inputs of Nets are coupled together such that it captures the dynamics of a stocks direction and momentum with a high degree of accuracy not being fooled by false reversals. This specialized neural net structure cannot be trained via normal back propagation methods. Only because this system was trained using Genetic Algorithms is this power available to the public. Specialize Neural Net structure considerably increases the capabilities of the Neural Net such that they are powerful enough to be PreTrained. No additional training in the field is required to utilize this system thus reducing required CPU resources tremendously. Which basically means you can run these nets on all of your investments simultaneously with minimal effect on CPU resources.
Purpose	For generation of real-time momentum and direction predictions to help optimize entry and exit points (this is a serious tool for day traders basically). Works best with high volume stocks and requires a minimum of 1 trade every 5 minutes to be moderately accurate. With high volume stocks the accuracy increases substantially and has been reported as the most accurate real-time neural net predictor currently available to the public (this was in comparison to similar programs that have this capability). Limitations are that it requires a real-time streaming data subscription from Prophet.Net and that the symbol is supported for streaming by Prophet.Net (which basically covers all US markets but unfortunately no Foreign markets at this time- if you would like to see Foreign markets supported in real-time streaming data please send an email requesting this feature in order to show Prophet.Net there are sufficient users wanting it).

Of course ANNI doesn't stop here since this is just the core details. There is much, much more thought put into carefully designing the systems ANNI offers in order to produce the highest level of outputs possible. That coupled with all the tools and integrations makes ANNI a must have investment program that you shouldn't make a single trade without. In short, simply trying ANNI is the best way to see for yourself what ANNI can do for you and your trading. Since there is a free, unrestricted 21 day trial available to everyone, there is absolutely nothing to lose. Go to the download page now to download your trial copy today and start a whole new way of trading.

<< Back | Tour Topics