Fig 1.
Decision consistency in CBL occurs both in photon systems and neural systems; we propose the local reservoir model for a common underlying model.
(A) Overall approach to the subject matter. (B) Decision making is coupled with an environment wherein the architecture is viewed by the relation among the “visible system” and the “local reservoir”.
Fig 2.
Local reservoir modeling for CBL consistency.
(A) A visible system, which acts as a physical system, is coupled with (B) a local reservoir in which environmental context exerts influence. Relaxation to the lower-left level in the visible system represents Decision L, which corresponds to an excitation in the upper-right level in the local reservoir. Conversely, relaxation to the lower-right level in the visible system denotes Decision R, which corresponds to an excitation in the upper-left level in the local reservoir. As a consequence, if the local reservoir accommodates excitations in an unbalanced manner, consecutively identical decisions can be observed in the visible system.
Fig 3.
CBL learning behavior via local reservoir.
(A) Decision consistency, which indicates the degree of CBL, exhibits higher values when the size of the local reservoir is small, whereas larger local reservoirs do not yield CBL. (B,C) Consecutive decisions are visualized in a random walk in the case of small and large local reservoirs. (D) The dependency to the internal dynamics of local reservoir (lifetime value). (E) Active portion of local reservoir as a function of the size of the local reservoir.
Fig 4.
Analytical modeling of local reservoir.
(A) The state transition diagram when the size of local reservoir (N) is 1; the number of lower energy levels is one. (B,C) When N = 2, the probability of making the same consecutive decision is higher than that of changing decisions when the internal dynamics of the local reservoir is slow.
Fig 5.
Single-photon-based system exhibiting CBL.
(A) Architecture for a decision-making system based on single photons. The angle of a linearly polarized single photon is configured by the half waveplate (λ/2). (B) The degree of precision (or resolution) of controlling the half waveplate. (C) The decision consistency, or CBL, exhibits larger values when the resolution is smaller, whereas it decreases as the resolution increases. (D) Active portion in the local reservoir as a function of the size of local reservoir.
Fig 6.
(A) Architecture for a decision-making system based on a cognitive model of the brain. (B) Decision consistency of CBL observed in human behavioral data (Occupation preference task; Nakao, et al. Sci. Rep. 2016 [10]) and the estimated size of local reservoir based on the model.