PROJECT TITLE :

Meta-Learning in Neural Networks A Survey

ABSTRACT:

In recent years, there has been a notable uptick in people's interest in the study of meta-learning, also known as learning how to learn. Meta-learning, on the other hand, seeks to improve learning algorithms themselves based on the experience of multiple learning episodes, in contrast to traditional approaches to artificial intelligence (AI), which start from scratch to solve problems using a predetermined learning algorithm. This paradigm presents an opportunity to address many of the traditional challenges that are associated with Deep Learning, such as the bottlenecks of data and computation, as well as generalization. The modern environment of meta-learning is analyzed and described in this survey. The first thing that we are going to do is talk about different definitions of meta-learning and where it stands in relation to other fields that are related to it, such as Transfer Learning and hyperparameter optimization. Following this, we present a new taxonomy for the classification of meta-learning methods that offers a more detailed breakdown of the available space. In this article, we take a look at some of the most successful and promising applications of meta-learning, such as few-shot learning and Reinforcement Learning. In conclusion, we talk about some of the outstanding challenges in the field as well as some of the promising areas for future research.