From White to Black through Gray... and one step back to Dark Gray

 

 

 

 

 

 

 

 

Aim and scope of the tutorial

 

 

 

 

 

Researchers and practitioners frequently need optimizers dedicated to solving various optimization problems. Choosing the optimizer appropriate for the considered problem may significantly improve the quality of the final solution. Moreover, if properly used, the knowledge about the problem's nature may lead to proposing results of outstanding quality and with a low computational cost.

 

This tutorial is dedicated to researchers and practitioners on all levels of expertise to those starting their work with optimization or wishing to know more about this fascinating branch of science and to those who are well-experienced in proposing and using various optimizers. To this end, we give a smooth introduction to the current research state in all considered areas. We focus on examples that show the nature of the observed problem features. Then, we clarify the intuitions behind the proposed optimizers and explain their motivations, pros and cons.

 

The tutorial is divided into four parts. First, we briefly discuss the issue of White-box optimization. In the second part, we focus on Gray-box optimization. We show its potential to speed up the efficiency of the optimizer significantly. However, we also discuss the details of the recently proposed Gray-box operators that improve the optimization quality. The third part focuses on Black-box optimization:

  1. We present the statistical-based problem decomposition techniques that triggered a breakthrough in optimizing many combinatorial problems.
  2. We introduce the empirical linkage learning techniques that enabled discovering the underlying problem structure with the quality close to the Gray-box-like (perfect) information.

 

The last part presents the Dark Gray optimization, which uses empirical linkage learning techniques to incorporate the Gray-box operators in Black-box optimization. Finally, we will show the new opportunities raised by such a fusion.

 

 

Tutorial length

1.5 hours

 

 

 

 

 

 

Tutorial level

introductory

 

 

 

 

 

 

Download