Abstract Image registration is a crucial and fundamental procedure in medical image analysis. Although many registration methods have been proposed, it is still a challenging task in some scenarios, such as images with large anatomical variations, multimodal registration, etc. Additionally, the scale and diversity of model imaging data have significantly increased, which pose more challenges for the registration algorithm. Machine learning techniques applied to image registration tasks can help address the aforementioned issues. Specifically, different machine learning techniques can be employed to learn from prior registration results to improve the registration performance in some challenging tasks. For instance, they can be employed for learning an appearance mapping model, learning an effective initialization for the optimization, etc. Recent studies have also demonstrated the potential of deep learning methods in addressing challenging registration problems. This chapter will be dedicated to summarizing state-of-the-art learning-based registration algorithms.