| 摘要 |
PROBLEM TO BE SOLVED: To solve the problem wherein cancer cells need to be classified quickly and efficiently and thus a method for classifying cancer cells into two types of ALL and AML is necessary. SOLUTION: A neural network is so configured that load factors provide a positive-by-positive multiplication for a desirable output, and a positive-by-negative multiplication to reduce an undesirable output. An input of an ALL group produces an ALL output larger than an AML output, and an input of an AML group produces an AML output larger than an ALL output. In the method that handles an input of normalized data to produce much the same positive/negative value, the load factor of the same sign as the input causes an excitation mode to increase the output, and the load factor of the opposite sign causes a reduction mode to reduce the output, so that the difference between both outputs is increased to improve identification performance. COPYRIGHT: (C)2004,JPO&NCIPI
|
