| 摘要 |
The invention relates to a self-organizing algorithm for mass analysis of gene expression data from DNA array experiments, which initially has a structure comprising two 'daughter' neurons connected to a 'mother' neuron, which consist of a list of values or profile, initially made up of random data. In a series of successive cycles, the algorithm divides the data set into subsets. Each of the cycles consists of a series of periods in which the data is presented to the terminal neurons, which are updated. When the updating carried out between two consecutive levels falls short of a given minimum level, it is considered that the network has converged in that cycle. It is then decided whether the network should continue to grow or has already completed its growth. If the latter occurs, the process or algorithm is ended.
|
