Bio-medical Ontologies Maintenance and Change Management

190 I.R. Godínez et al.
On the other hand, the work of Lukashin et al. [17] is one of the earliest
researches done in promoters localization in DNA sequences by using artificial
neural networks. The process therein mentioned uses a small subset (around
10%), taken randomly from the whole set of sequences of promoters, in order
to train the network. The learning capacity is tested by presenting the whole
set of sequences, obtaining an effectiveness of 94% to 99%.
With respect to gene localization, quantitative analysis of similarity between
tRNA gene sequences have been done, mainly for the search of evolutive
relationships. New sequences have been introduced to this neural network and
these are known as tRNA genes [18].
Another example related to promoters localization is a multi-layer feedforward
neural network whose architecture is trained in order to predict
whether a nucleotide sequence is a bacterial promoter sequence or not [19].
2.2 Alpha-Beta Associative Memories
Basic concepts about associative memories were established three decades
ago in [4], [5], [21], nonetheless here we use the concepts, results and notation
introduced in the Yáñez-Márquez’s Ph.D. Thesis [22]. An associative memory
M is a system that relates input patterns and outputs patterns, during
the operation, as follows: x → M → y with x and y the input and output
pattern vectors, respectively. Each input vector forms an association with a
corresponding output vector. For k integer and positive, the corresponding
association will be denoted as: � x k ,y k� .
Associative memory M is represented by a matrix whose ij-th component
is mij. MemoryM is generated from an apriorifinite set of known associations,
known as the fundamental set of associations. Ifμ is an index, the
fundamental set is represented as:
{ (x μ ,y μ ) | μ =1, 2,...,p }
with p the cardinality of the set. The patterns that form the fundamental set
are called fundamental patterns.
If it holds that x μ = y μ , ∀μ ∈{1, 2,...,p }, M is autoassociative, otherwise
it is heteroassociative; in this latter case it is possible to establish that
∃μ ∈{1, 2,...,p} for which x μ �= y μ . A distorted version of a pattern x k
to be recalled will be denoted as ˜x k . If when feeding a distorted version of
x ϖ with ϖ = {1, 2,...,p} to an associative memory M, it happens that the
output corresponds exactly to the associated pattern y ϖ , we say that recall
is correct.
Among the variety of associative memory models described in the scientific
literature, there are two models that, because of their relevance, it is
important to emphasize: morphological associative memories which were introduced
by Ritter et al. [6], and alpha-beta associative memories, which
were introduced in [22]-[24]. Because of their excellent characteristics, which
allow them to be superior in many aspects to other models for associative