The Pattern Recognition Problem

The shape-space move toward proposes as an attribute string s = ? s1, s2, . . . , sL ?in an L dimensional shape-space, S, (s ∈ SL), can show any immune cell or molecule. All attributes of this string is invented to show a feature of the immune cell or molecule, as like: its charge, Van Der Wall interactions. In the improvement of artificial immune systems the mapping from the attributes to their biological counterparts is frequently not applicable. The kind of attributes utilized to represent the string will explain partially the shape-space beneath study, and it is highly dependent upon the problem domain. Any of shape-space constructed from a limited alphabet of length k constitutes a k-ary Hamming shape-space. As an illustration, an attribute string built on the set of binary elements {0, 1} consequent to a binary Hamming shape-space. This can be consideration of, in this case, like a problem of recognizing a set of characters shown by matrices composed of 0's and 1's. All elements of the matrixes correspond to pixels in the character. We have Euclidean shape-space, if the elements of s are shown by real-valued vectors. Most of the artificial immune systems   found in the literature employ binary Euclidean or Hamming shape-spaces. Other kinds of shape-spaces are possible also, as like symbolic shape-spaces that combine various symbolic or attributes in the show of a single string s. These are generally found in data mining applications, whereas the data might have symbolic information as like age, name of a set of patterns.

Another significant characteristic of the artificial immune systems or AIS is that most of them are population based. This means that they are composed of a set of individuals, showing immune molecules and cells that have to perform a described role; in our clonal selection, negative selection, and immune network, each of them rely on a population M of individuals to recognize a set P of patterns. The negative selection algorithm have to be explain a set of detectors for non-self patterns; clonal selection reproduces, maturates, and

Chooses self-cells to recognize a set of non-self; and the immune network keeps a set of individuals, linked as a network, to recognize self and non-self.

Suppose now the availability of a set of N patterns or antigens p_i, i = 1, . . ., N (p_i ∈ P) to be recognized, and a set of M immune cells or and molecules or antibodies m_j = 1, . . . , M (m_j ∈ M) to be utilized as pattern recognizers via negative, clonal or network algorithms. Suppose, that both have the similar length L also as (p_i, m_j ∈ SL).

Let first the binary Hamming shape-space case that is the most broadly utilized. There are some expressions such can be employed in the finding of the degree of match or affinity between an element of M and an element of P. The simplest case is to only compute the Hamming distance or DH between these two elements, as specified by Eq. 1.

Other approach is to search for a sequence of r-contiguous bits [13], and a specified threshold, after recognition is said to have happen, if the number of r-contiguous matches among the strings is greater. This final method has the advantage that this favours sequences of complementary matches, hence searching for related regions among the attribute strings or patterns.