1: \begin{abstract}

2: In this paper, we propose a novel technique to implement stochastic gradient methods, which are beneficial for learning from large datasets, through accelerated stochastic dynamics. 

3: A stochastic gradient method is based on mini-batch learning for reducing the computational cost when the amount of data is large. 

4: The stochasticity of the gradient can be mitigated by the injection of Gaussian noise, which yields the stochastic Langevin gradient method; this method can be used for Bayesian posterior sampling. 

5: However, the performance of the stochastic Langevin gradient method depends on the mixing rate of the stochastic dynamics. 

6: In this study, we propose violating the detailed balance condition to enhance the mixing rate. 

7: Recent studies have revealed that violating the detailed balance condition accelerates the convergence to a stationary state and reduces the correlation time between the samplings. 

8: We implement this violation of the detailed balance condition in the stochastic gradient Langevin method and test our method for a simple model to demonstrate its performance.

9: \end{abstract}

10: