Qubit Neural Network and Its Efficiency Noriaki Kouda, Nobuyuki Matsui, Haruhiko Nishimura, and Ferdinand Peper Abstract Though neural networks have attracted much interest in the last two decades for their potential to describe brain function realistically, they have failed thus far to provide models that can be simulated in a rea- sonable time on computers, other than toy models. Quantum computing is a likely candidate for improving the computational efficiency of neural networks, since it has been very successful in doing so for a selected set of computational problems. In this framework, the Qubit neuron model, proposed by Matsui and Nishimura, has shown its promise in improving the efficiency. The simulations we have performed have shown that the Qubit model performs learning problems with significantly improved effi- ciency as compared to the classical model, and a classical model in which complex numbers are allowed as model parameters and as input. In this paper, we investigate what contributes to the efficiencies through 4-bit parity check problem known as a basic benchmark test. Our simulation suggests that the improved performance is due to the use of superposi- tion of neural states and the use of the probability interpretation in the observation of the output states of the model.