| 摘要 |
PROBLEM TO BE SOLVED: To use only necessary modules by providing a means which regards a module as a Fermi particle system space in a low-temperature grand canonical state and a means which regards an input signal as a parameter representing an energy track obtained by observing Fermi particles in the Fermi particle space. SOLUTION: The learning device 1 is composed of modules 10, 20, and 30 and a decision means 40 which calculates weights by the modules according to information from the modules 10, 20, and 30. The modules 10, 20, and 30 are provided with input signal conversion parts 11, 21, and 31, energy track determination parts 12, 22, and 32, presence establishment determination parts 13, 23, and 33, and learning parameter updating parts 14, 24, and 34. The decision means 40 is provided with an observation space determination part 41 and a free energy determination part 42. In this case, the modules 10, 20, and 30 are in a low-temperature grand canonical state. Further, the input signal is a parameter of the energy track of an observed Fermi particle. In this constitution, a memory can effectively be used.
|
