Meta-Learning

Meta-Learning: An Overview explains the fundamentals of meta-learning, giving an understanding of the concept of learning to learn. After giving a background to artificial intelligence, machine learning, deep learning, deep reinforcement learning, and meta-learning, it provides important state-of-the-art mechanisms for meta-learning, including memory-augmented neural networks, meta-networks, convolutional siamese neural networks, matching networks, prototypical networks, relation networks, LSTM meta-learning, model-agnostic meta-learning, and Reptile. It then demonstrates the application of the principles and algorithms of meta learning in computer vision, meta-reinforcement learning, robotics, speech recognition, natural language processing, finance, business management, and health care. A final chapter summarizes the challenges, opportunities and future trends.Meta-Learning: An Overview gives students and researchers and understanding of the principles and state-of-the-art meta-learning algorithms, enabling the use of meta-learning for a range of applications. A comprehensive overview of state-of-the-art meta-learning techniques Presents the three approaches to meta-learning: model-based, metric-based and optimization-based Gives strategies of meta-learning in multiple subfields of machine learning and AI that focus on developing versatile systems, including unsupervised learning, Bayesian inference, multi-task learning, transfer learning, and lifelong learning Presents applications in computer vision, meta-reinforcement learning, robotics, speech recognition, natural language processing, finance, business management