## Optimizasyon | IOSO

### IOSO Optimizasyon Yazılımı

Sigma Technology, www.iosotech.com

IOSO_NM_info.pdf

IOSO_PM_info.pdf

IOSO is a novel optimization strategy based on a new generation optimization technology.

IOSO stands for Indirect Optimization based on Self-Organization. It is used to improve the performance of complex systems, technical facilities and technological processes and to develop new materials based on a search for their optimal parameters.

Without the use of specialized optimization tools, the designer encounters an infinite number of combinations of project variables and is forced to substantially limit the scope of optimal values to be searched for, which actually has a negative impact on the output.

IOSO enables the designer to easily integrate all computation models or simulation applications (commercial as well as in-house) into a single computation unit and automate the optimal solution search process based on the IOSO optimization technology embedded in IOSO software products.

**“We assume that mathematical models and software packages for creating objects or performing model calculations (however full and accurate they may be) is not a sufficient condition for successful design and modification of modern technical and other systems. Creating competitive samples requires integration of mathematical models, modeling applications or a real-life object with search methods of study as part of a single ‘optimization environment’. This environment is what we call optimization technology.”**

Prof.Igor N.Egorov (the chief developer of IOSO technology)

Third-party references to IOSO

George S. Dulikravich, Florida International University

AFOSR - Air Force Office of Scientific Research, US, (Page 2):

* Currently, a Russian commercially available software named IOSO is the most efficient and the most robust multi-objective optimization software… IOSO, which involves concepts of neural networks, radial basis functions, and self-adapting response surface methodologies, requires the minimum number of the objective function evaluations and that is the most versatile and robust multi-objective optimizer. *Details…

Timothy W. Simpson, The Pennsylvania State University, US

Vasilli Toropov, University of Leeds, UK, (Page 13):

*. Details...*

**IOSO offers unique state of the art optimization algorithms that are based on self-organizational strategy and efficiently combine traditional response surface methodology with gradient-based optimization and evolutionary algorithms in a single run. The offered algorithms are equally efficient for the problems of complex and simple topology that may include mixed types of variables**Carlos A. Coello Coello and Ricardo Landa Becerra

Evolutionary Computation Group Departamento de Computación, Mexico, (6.5 Design of alloys):

*. Details...*

**IOSO consists of two stages. In the first stage, an approximate model of the objective functions is created. In the second stage, this approximate model is optimized. IOSO incorporates evolutionary algorithms, and artificial neural networks with radial basis functions that are used to build the response surfaces. The idea is to use this metamodel (or approximate model) to perform a very reduced number of evaluations of the actual objective functions of the problem**

IOSO approach to Optimization

IOSO is a universal optimization strategy. IOSO features engineers-focused approach and could be used by an ordinary engineer having almost nothing theoretical knowledge in optimization. The efficiency of standard optimization algorithms while applying on each concrete task depends strongly on the values of the number of its settings (they may have a dozen of parameters), inadequate settings strongly limits the efficiency of these algorithms in practice. Moreover any standard algorithm that included in almost any commercial optimization package is designed for the successful solution of some special optimization problems (with the appropriate topology of optimization function in terms of mathematics). To successfully apply it an engineer needs to know what the best algorithm to choose for such special optimization task (this is so-called the choice of optimization strategy). IOSO technology in contrast features automatic adaptation procedure of internal parameters; there is no need to find optimization strategy using IOSO technology.

Scalability of IOSO. Most of standard optimization techniques rapidly loose their efficiency with the increase of the scale of optimization tasks. They show quite good results on small-scale or medium-scale optimization tasks (with the number of independent parameters ~ 10 - 30) but for large-scale real-life optimization tasks of the order of 50 – 100 parameters and more it is a well-known fact that the efficiency of these methods is very low (they are too time-consuming). That was the one of the main reasons why the group of Prof Egorov started their own developments in optimization, because in the field of high-tech engineering, for example, Turbomachinery or Aerospace they need to solve large-scale optimization tasks of the order of 50 - 100 design variables and more and traditional approaches couldn’t handle them efficiently. IOSO feature scalability properties having good efficiency on small, medium and large-scale optimization tasks.

The basic structure of IOSO technology may be explained in comparison with traditional methods like in the picture below

### IOSO technology features two main components:

a) It permanently builds local and global approximation functions of complex (self-organized) evolutionary structure and optimization algorithms work mostly with approximation functions (or surrogate models) other than mathematical models. This sufficiently reduces the number of direct calls to mathematical model and reduces the time required for optimization.

b) It applies original moving strategy (search region adaptation) that leads local search regions to global extreme and it is achieved with very high probability.