| 摘要 |
PURPOSE:To obtain a stable and a high inference accuracy by carrying out a learning of a learning vector quantization member by making an electromagnetic field distribution by imaginary dipoles as learning data, and inferring the number of dipoles of an organism or an object by using the result of the learning. CONSTITUTION:When dipole data which consists of the number, the positions, and the moments of dipoles is inferred from the electromagnetic field distribution on the head skin, in case of generating a single or plural dipoles inside a human brain, the dipole data 1000 which consists of the number, the positions, and the moments of the dipoles is produced at a dipole data producer 100, and it is delivered to a regular problem operation unit 101. In the operation unit 101, form information of the head is obtained, the electromagnetic field distribution generated at the parts of the surface of the head is calculated, and the set of electromagnetic field data and the dipole data 1000 is made into learning data 1001, which is delivered to a data selector 102 for vector and a learning vector quantization member 103. And in the quantization member 103, the learning is carried out by the algorithm of the learning vector quantization so as to obtain the electromagnetic field data, and the dipole data is inferred. |
