The aim is to perform to realize a learning system for active perception using a neural network. It obtains inputs only from a movable visual sensor and learns both appropriate recognition and sensor motions for effective perception. The proposed learning method is based on reinforcement learning using reinforcement signals calculated from the recognition results. We conclude from simulations that it enables the system to move a visual sensor to the appropriate location and finally classify presented patterns correctly. The recognition rate was better than that in a simulation for comparison where the sensor was fixed at the initial location.
The cholinergic input to the hippocampus originates in the septum and diagonal band. A bath application of carbachol, a cholinergic agonist, induced different patterns of rhythmical waves depending on its concentration in guinea pig hippocampal slices. Tetanus-induced long-term potentiation (LTP) was differentially facilitated under several concentrations of carbachol. Under the concentrations, which were within the optimum range for generating theta-like waves, LTP was most facilitated. In addition, the higher augmentation of LTP occurred when LTP was induced during the generation of theta-like wave. These results suggested that depending on the activity of the cholinergic septal input to hippocampus, different patterns of rhythmical waves are induced and the plasticity in hippocampal neural networks can be modulated by them.
The existence of spurious memories is a serious problem in associative memory systems based on recurrent neural networks. We propose an associative memory system which consists of hysteresis neurons and pseudo-inverse matrix weights, and show that it has no spurious memories. First, we show the patterns which are not included in the space spanned by the stored patterns, namely memory space, are unstable in dynamics of hysteresis neurons using the characteristics that pseudo-inverse matrix weights project the input vector onto the memory space. Next, we consider the condition under which no patterns except for the stored ones exist in the memory space.