Artificial neural networks have actively researched as future electronics technologies. However, since the Neumann-type arithmetic processing method is not efficiently capable to process a large amount of information, a neurocomputing method specialized for parallel computing has recently received much attention to solve the problem. We fabricated the neuromorphic device by depositing Ga-Sn-O (GTO). an oxide semiconductor that can be highly integrated by fonning a thin film, on the LSI. We evaluated the synaptic characteristics of GTO thin films and confirmed the basic operation of synaptic devices. It was also confirmed that our device was able to recognize letters.