The analysis of a trained deep convolutional neural network (DCNN) model enhances our understanding of both the artificial methodology for information processing in object classification and the neural mechanisms underlying visual object perception. The AlexNet model based on DCNN for object classification progressively organizes artificial representations through its network hierarchy similar to the neural system for object perception. Previous studies have indicated that the obtained mechanism of a trained AlexNet model using ImageNet dataset might, at least in part, mirror the ventral stream in primates for visual object recognition. However, the above results might be specific to ImageNet dataset. To examine the universality of the results, here, we trained AlexNet using alternative datasets, CIFAR-10 and investigated the relationships between the artificial representations in layer of a CIFAR-10-trained AlexNet model and the neural representations in the primary visual (V1), intermediate visual (V4), and inferior temporal (IT) cortices. Similar to the ImageNet-trained AlexNet, the artificial representations in the convolutional layers of the CIFAR-10-trained AlexNet model markedly correlated with neural representations observed in V1 and V4 cortices. Conversely, across all levels of the visual cortices, the artificial representations within the fully connected layers of a CIFAR-10-trained AlexNet model markedly differed from the neural representations. This finding is distinct from previous studies suggesting that the model neuron responses in the fully connected layers of an AlexNet model trained with ImageNet dataset exhibit similarities to the neuronal responses in V4 and IT.