~Russian Bear’Z Blog~

GA4TS – is a dynamic library set aside for the optimization of strategies using one or multiobjective in TradeStation Omega (any version). The advantages of the algorithm embodied in GA4TS are:

1. The possibility of searching multiobjective optimal solution (using Pareto Set).
2. The search is carried out with a help of Genetic Algorithm.
3. Greater speed of convergence in comparison with other methods (GA).
4. High accuracy of all calculations.

The library includes:

1. The library allows you to use all the opportunities of the algorithm.
2. The opportunity of the utilization by means of special interface in all versions of Omega TradeStation.
3. The opportunity of tuning and correction of the parameters of the reproductive cycle in the Genetic Algorithm.
4. The addition of good and well-known decisions into the basic population.
5. The bitwise accuracy of all calculations.
6. The filtration of data while outputting it.

The use of GA is justified in such cases when an enormous calculating capacity is required. And quick convergence allows you to solve the tasks amounting 10⁵⁰ in a several hours. The speed of convergence of this method is greater in comparison with other realizations of GA. As used here, the speed is understood to be the quality of the iterations done (calculations of one element or a strategy with specific parameters).

For that let s consider an example from MATLAB Rastrigin’s Function fig.1 in the description to “Genetic Algorithm and Direct Search Toolbox”. The function is:

Ras(x,y) = -(x²+y²)+10(-2 + cos(2x) + cos(2y)),

in which all the identical transformations aimed at presentation of calculations accuracy are done, as 1 +10^-16 on PC is 1 as a rule.
Fig. 1. Rastrigin’s Function.

Continuation:

Apparently the maximum of this function is reached at 0 (at the point with position data (0;0). For the definitional domain [-1,1]x[-1,1] GA Toolbox has solved this task at the size of population 100 and using 100 generations: with the result 0.000030 at the point (0.000150, -0.000400).Consequently approximately 10 000 iterations were used. With the help of GA4TS was obtained the result with the established exactness 10^-20, the value of which is -6.06e-021 at the point (-1.90e-011, -7.54e-011), less than for 7000 iterations.

Fig. 2. Convergence of results.

The dynamics of the results Rastrigin s Function signed plus (for the reflection on the logarithmic scale) is shown on pic. 2: The values of all iterations are above and the best values for each generation - below.

Also for this purpose was obtained a result up to 30 signs taking into account that not restricting the work of GA by time constrains in MATLAB we have not succeeded in achieving the result up to 8 sign.

It is clear that such accuracy is not very often essential in the practical tasks, but in the experiment the conclusions concerning the convergence of the algorithm are made on this ground.

Let s view one more task: under the same conditions for 2 objective (Ras(x,y), x²+y²), less than 8000 iterations , was found a Pareto Set , including the solution (-0.000062, ·), (·,1.998541), lower and upper are shown on figure 3.In such statement only one lowermost decision answers, previous, one-objective to a problem that can hide very interesting other decisions.

Fig. 3. Pareto Set, including the solution Ras(x, y) = -0.000062.

That is, approximately for the same quantity of iterations were solved the tasks of various complexity, using genetic algorithms that differ in complication.

The scope of the library GA4TS is described in details in the instruction.

The library can be got for 1890 USD,
electronic or bank payment.
Feedback.

Article on a theme:
Search of the optimal strategies using Pareto Set

---