Nature has proved to be the best testing system, where we can analyze the effectiveness of any method of solving problems. It provides one of the most complex problems to be resolved: the survival. Analyzing how the species behave to achieve that survival, soft computing methods try to mimic this behavior to provide meaningful solutions to diverse problems. This chapter offers an introduction the fundamentals that the different soft computing techniques translate from Nature. It includes an approach of the brain behavior (Artificial Neural Networks) or the evolution ideas taken from Darwin’ laws (Evolutionary Computation algorithms).
TopArtificial Neural Networks
As mentioned above, perhaps the supremacy of human over all the other living creatures is due to his high intellectual capacity and the basis of this capacity lies in the brain. However, reproducing the overall performance of the brain in order to achieve a problem solving system is not feasible at all.
The brain provides action responses to the whole range of stimuli received from the outside world: images, sounds, tastes, smells, temperatures, etc. To this end, it simply uses its collection of neurons (about 1011) and its many synaptic connections between them (from about 1,000 to 10,000 synaptic connections can be found in one neuron). These scales are impossible to reach in a simulation system, but there is a possibility of building more simple models based on each of the major components of the above-mentioned performance: the neurons and interconnection architectures between them.
Both elements make up the basic structure of an ANN. Therefore, an ANN may be considered an attempt to produce learning systems inspired by nature (based on abstract models of how we think and how the brain works).
Biological Basis
Although the outline of the full performance of neurons is still to be discovered, we have quite a good understanding of the whole process.
Broadly speaking, a neuron receives through its extension branches, called dendrites, a series of impulses from its neighbouring neurons. Depending on the overall final intensity of the stimuli, there may be an excitatory or inhibitory effect on the neuron. In the first case, the neuron will cause an output signal which is transmitted through its axon to the dendrites of the neighbouring neurons. Information is stored by means of this exchange of impulses and in the capacity of neurons to be activated or inhibited by a certain set of inputs. Obviously this process occurs at the same time in millions of neurons, so the most fundamental characteristics of the brain: high redundancy, fault tolerance partial adaptability, and so on, are a consequence of this decentralized behaviour.