| 摘要 |
PURPOSE:To improve considerably the storage efficiency by dividing each memory vector into plural subvectors and storing the own correlation and the mutual correlation among the divided vectors. CONSTITUTION:A partial vector arithmetic part divides an input picture y1 into partial vectors. The picture is divided into 16 equal parts in this example. In this case, the picture yI is obtained as shown by the equation. Here yI(q) shows a vector which uses all picture element values as the component elements within the q-th partial picture. A correlation arithmetic part 102 supplied partial vectors yI(1)-yI(16) to calculate a correlation matrix. A memory matrix arithmetic part 103 supplies these correlation matrices and also extracts the elements of the partial correlation matrices showing the pictures so far stored to replace them. Then an input X is applied to retrieve the memory information, and a vector arithmetic part 201 fetches the remembrance input and divides it into 16 partial vectors. Receiving these partial vectors, a remembrance arithmetic part 202 performs a desired operation to obtain an entire image of a storage item closest to the input X. |
