You're working on a deep learning model that has a small number of features but a large number of instances. Despite your efforts, the model's performance is below expectations, and you suspect it's underfitting. Which technique would you consider to enhance its performance?