During the process of batch training a neural network, you observe that the training loss shows oscillatory behavior, indicating instability in the learning process. This could prevent the model from converging to an optimal solution. What modification should you apply to your model's hyperparameters to help it converge more reliably?