The learning property in size-variable Boolean nets

We report in this paper the construction of a size-variable Boolean net and the investigation of the learning property. The net is constructed so that it allows a change in size with the increase of the input-output cases to be learned. Adaptability to the dynamic environment is considered to be the characteristic flexibility of living organisms. We have constructed a learning model in which the Boolean net size can be increased, and the adaptation is realized as the net size is increased. The learning model is more efficient compared with fixed size nets. However, the flexibility of living organisms should not be expressed as fixed operations, but should be considered to include the deviation from the global tendency of adaptation. A reverse operation is introduced to realize and to show the deviation. We attained higher learning velocity with the learning model having the reverse operation.