Repeated exposure to a specific stimulus can enhance animal's sensitivity to it so that the perceptual capability is improved. This experience-induced perceptual improvement is referred to as perceptual learning. However, the neural system has some robustness and is not necessarily modified by its any input. In the case of visual perceptual learning (VPL), perceptual performance for a task-relevant stimulus can be selectively improved without any sensitivity change to task-irrelevant stimuli which are presented even simultaneously with the task-relevant one. In this study, we propose a feed-forward spiking neural network model consisting of a primary visual cortex (VI) layer and a higher visual area (V4) layer; their inter-layer feed-forward connections are modified by synaptic learning in a particular interest in how VPL can be affected by neural activities in the higher area due to attentional signals. Through simulations, we show attentional inputs are needed to facilitate inter-layer synaptic learning which yields improved sensitivity to the task-relevant stimulus, and thus to increase the task performance.