The adaptation challenge in machine learning and deep reinforcement learning is to learn from past learning experiences to address new tasks. Despite the importance of recursion in learning, most programming synthesis models still focus on training models to solve specific tasks.