Changes between Version 40 and Version 41 of GAIIA

Timestamp:

Jan 12, 2010, 11:16:11 AM (14 years ago)

Author:

mihai.istin

Comment:

--

GAIIA

@@ -41 +41 @@
  * minimum idle times 
 
-The proposed algorithm
-    Our solution is based on a hybrid evolutionary algorithm with combines the benefits of genetic and immune algorithms algorithms in order to provide efficient solutions. It is known that the convergence time of genetic algorithms is highly influenced by the average fitness of the initial population. So if we could find a way to provide a population with an good average fitness at the initialization phase of the genetic algorithm, the algorithm will converge faster. 
-   Immune algorithms are also bio-inspired algorithms that are based on the human immune system model. They evolve a population of antibodies in order to fit better the antigens that are threatening the immune system. 
-   As a computational model, immune algorithms can be designed to evolve a population of chromosomes representing the antibodies in order to have a good fitness. The implementation of such algorithms is based on the clonal selection principle which model the response of an immune system to an intruder. Basically, the current population of antibodies is evaluated and the best individuals (individuals with the best fitness) are selected for the maturation process. During this process, for each selected individual will be made a number of clones proportionally with its fitness. Then each clone will suffer multiple mutations. The number of mutations will be inverse proportionally with its fitness. Then, the clones are evaluated, and the best are selected in order to survive to the next generation. The antibodies with the lowest fitness will be removed from the current population. 
-   The main advantage of immune algorithms is that they are capable of evolving a population with a good average fitness after only several iteration. On the other hand these algorithms have also a drawback: as a consequence of the fact that the mutation rate varies inverse proportionally with the fitness, the best individuals will be similar. But if we are aware of this drawback we can control the diversity of the evolved population by using only a few generations and by using reinsertion.  
- * uses immune algorithms
+
+== The proposed algorithm ==
+
+   
+
+Our solution is based on a hybrid evolutionary algorithm with combines the benefits of genetic and immune algorithms algorithms in order to provide efficient solutions. It is known that the convergence time of genetic algorithms is highly influenced by the average fitness of the initial population. So if we could find a way to provide a population with an good average fitness at the initialization phase of the genetic algorithm, the algorithm will converge faster. 
+   
+
+Immune algorithms are also bio-inspired algorithms that are based on the human immune system model. They evolve a population of antibodies in order to fit better the antigens that are threatening the immune system. 
+   
+
+As a computational model, immune algorithms can be designed to evolve a population of chromosomes representing the antibodies in order to have a good fitness. The implementation of such algorithms is based on the clonal selection principle which model the response of an immune system to an intruder. Basically, the current population of antibodies is evaluated and the best individuals (individuals with the best fitness) are selected for the maturation process. During this process, for each selected individual will be made a number of clones proportionally with its fitness. Then each clone will suffer multiple mutations. The number of mutations will be inverse proportionally with its fitness. Then, the clones are evaluated, and the best are selected in order to survive to the next generation. The antibodies with the lowest fitness will be removed from the current population. 
+
+
+The main advantage of immune algorithms is that they are capable of evolving a population with a good average fitness after only several iteration. On the other hand these algorithms have also a drawback: as a consequence of the fact that the mutation rate varies inverse proportionally with the fitness, the best individuals will be similar. But if we are aware of this drawback we can control the diversity of the evolved population by using only a few generations and by using reinsertion. 
+ 
+
+==  The Immune Algorithms ==
+
+In the following figure is presented the general model for the immune algorithms. The number of clones is computed using the following formula 
 [[Image(ia.jpg)]]
  * uses genetic algorithms