We were an early entrant within the financial services industry in applying advanced mathematics, statistics and Artificial Intelligence (A.I.) to market analysis and money management.

Artificial Intelligence is one of the fastest growing areas of market research on Wall Street Today. It is computer assistance in tasks requiring human reasoning, intelligence, intuition and imagination. Using knowledge-based "expert" systems , neural networks and genetic algorithms , this technology is capable of resolving problems of uncertain outcome that normally require human judgment and experience.

Properly programmed, the results of such problem solving efforts are accurate within a very narrow range of tolerance or confidence. Looking at today's market place, with its thousands of inter-relationships and increasing number of variables that continue to grow exponentially, A.I. is revolutionizing the way we approach market analysis.

An A.I. decision support system can do mechanically what a team of skilled market analysts would do. What makes it better, is that it can analyze more data faster and more reliably than a human can and do it without bias or emotion. The most crucial investment decisions are made at market turning points. That is where human emotion gets in the way of judgment. The computer eliminates that problem.

A.I. systems offer distinct advantages over traditional methods of analysis. They have the ability to efficiently process thousands of pieces of data and consider millions of possible combinations to discover relevant relationships that may not otherwise be apparent. Such analysis is extremely rapid, unemotional, and unaffected by personal bias. These systems are dynamic and self adapting. Unlike conventional market analysis that relies on static approaches, these new methods are able to adjust to changing market conditions and downplay or eliminate factors that are no longer relevant.

The LBS Systems provide a dynamic, self adapting decision mechanism that is intelligently sensitive to changing market conditions. The process is disciplined, completely objective and highly risk-averse.

Expert Systems

Expert systems are computer programs designed to review a set of facts (market conditions) and apply a set of rules (knowledge base) to arrive at the same conclusion that a team of experts would make if presented with the same set of facts. There are two primary components to an expert system: knowledge base, and inference engine.

An expert system begins by capturing human expertise in the form of IF/THEN rules. For example, the rule "IF the traffic light is red THEN step on the brake," captures one small aspect of driving expertise. Often dozens (or hundreds) of such rules are required to arrive at solutions to complicated, financial market problems.

The next step in the process involves applying the rules against a set of facts to determine which rules apply. Embedded within the expert system is an inference engine which determines when and how to apply the rules. The unique order in which the rules are applied to a given set of facts allows the expert system to be dynamic. As one might expect given the features and capacity of these technologies, the number of potential good solutions to a problem can be vast. The inference engine helps to quantify the solution sets to provide our portfolio managers with consistent, unbiased, unemotional problem analysis.

Neural Networks

Neural networks, also referred to as neural nets, mimic the human brain's most powerful ability, that of pattern recognition. A neural net can make a decision or a prediction by drawing on an inventory of patterns previously learned, seeking the one most relevant to the current situation. What distinguishes neural nets from "classical" Artificial Intelligence paradigms, such as rule-based systems, is that they "learn" by example and modify themselves, rather than having to be programmed with specific preconceived rules.

Neural nets use layers of real or "simulated" electronic neurons with various numeric weights. These weights are dynamically modifiable in the "training" phase representing the connection strengths between them to "learn" a predictive ability. A trained neural net can be described as a matrix of neurons and connection strengths that represent the "solution" to a problem. Commercially available neural net shells, as well as those still being researched, use a variety of algorithms to adjust the connection strengths between the neurons until convergence is achieved and the system is able to make correct predictions with a significant percentage of accuracy. Once the network successfully sets the weights of the connections to produce the correct output, the trained net is ready to use as a predictive or diagnostic tool.

The net can then be used by the portfolio manager to make predictions of future market action or to categorize the behavior of the current market.

Genetic Algorithms

Genetic algorithms are general-purpose search techniques for solving complex problems. Based upon genetic and evolutionary principles, GAs work by repeatedly modifying a population of artificial structures through the application of selection, crossover, and mutation operators. The choice of an appropriate structure for a particular problem is a major factor determining a GAs success. GAs have traditionally been used for optimization, but with a few enhancements, can be applied to classification and prediction as well.

We first examine how a traditional GA performs optimization. The goal in optimization is to find the best possible solution to a problem. In real-world problem solving, one does not usually know the best possible solution. Therefore, a more realistic objective is to find a good solution; or, given a current benchmark, to search for a better solution. A GAs fitness function measures the quality of a particular solution.

The traditional GA begins with a population of n randomly generated structures, where each structure encodes a solution to the task at hand. The GA proceeds for a fixed number of generations. During each generation, the GA improves the structures in its current population by performing selection, followed by crossover, followed by mutation.

• Selection is the population improvement or "survival of the fittest" operator. Basically, it duplicates structures with higher fitnesses and deletes structures with lower fitnesses.
• Crossover, when combined with selection, results in good components of good structures combining to yield even better structures. The offspring are the results of cutting and splicing the parent structures at various crossover points.
• Mutation creates new structures that are similar to current structures. With a small, prespecified probability, mutation randomly alters each component of each structure.

TERMS AND CONDITIONS
Please note, performance is not guaranteed, and past performance should not be taken as any indication of future performance.

Actual performance may vary, and the possibility of loss always exists.