Although noradrenaline and adrenaline are some of the most important neurotransmitters in the central nervous system, the effects of noradrenergic/adrenergic modulation on the striatum have not been determined. In order to explore the effects of adrenergic receptor (AR) agonists on the striatal firing patterns, we used optogenetic methods which can induce continuous firings. We employed transgenic rats expressing channelrhodopsin-2 (ChR2) in neurons. The medium spiny neuron showed a slow rising depolarization during the 1-s long optogenetic striatal photostimulation and a residual potential with 8.6-s half-life decay after the photostimulation. As a result of the residual potential, five repetitive 1-sec long photostimulations with 20-s onset intervals cumulatively increased the number of spikes. This 'firing increment', possibly relating to the timing control function of the striatum, was used to evaluate the AR modulation. The β-AR agonist isoproterenol decreased the firing increment between the 1st and 5th stimulation cycles, while the α-AR agonist phenylephrine enhanced the firing increment. Isoproterenol and adrenaline increased the early phase (0-0.5s of the photostimulation) firing response. This adrenergic modulation was inhibited by the β-antagonist propranolol. Conversely, phenylephrine and noradrenaline reduced the early phase response. β-ARs and α-ARs work in opposition controlling the striatal firing initiation and the firing increment.

Previous studies have shown that some combo of human cognitive biases is effective in machine learning. The well-used model of the biases is called loosely symmetric (LS) model. We show the efficiency and accuracy of our loosely symmetric model and its implementation of two cognitive biases, symmetry and mutual exclusively.In this study, we use loosely symmetric as a binary classifier to enhance its accuracy in small datasets.

No abstract

No abstract

SummaryAs the scope of reinforcement learning broadens, the number of possible states and of executable actions, and hence the product of the two sets explode. Often, there are more feasible options than allowed trials, because of physical and computational constraints imposed on the agents. In such an occasion, optimization procedures that require first trying all the options once do not work. The situation is what the theory of bounded rationality was proposed to deal with. We formalize the central heuristics of bounded rationality theory named satisficing. Instead of the traditional formulation of satisficing at the policy level in terms of reinforcement learning, we introduce a value function that implements the asymmetric risk attitudes characteristic of human cognition. Operated under the simple greedy policy, the RS (reference satisficing) value function enables an efficient satisficing in K-armed bandit problems, and when the reference level for satisficing is set at an appropriate value, it leads to effective optimization. RS is also tested in a robotic motion learning task in which a robot learns to perform giant-swings (acrobot). While the standard algorithms fail because of the coarse-grained state space, RS shows a stable performance and autonomous exploration that goes without randomized exploration and its gradual annealing necessary for the standard methods. 1.