Machine Learning for End Consumers

Abstract

The articles in this special section examine machine learning (ML) for end consumers. ML is a discipline that grew out of artificial intelligence (AI). At a minimum, an intelligent agent needs to perceive the environment around it, deliberate, and take the best course of actions to maximize some actual or estimated performance measures. ML was originally a trait of AI that concerned training intelligent agents to perform tasks that cannot be preprogrammed. ML has received much attention recently with advances in technologies that permeate many facets of our everyday lives, e.g., autonomous vehicles, lifelike chatbots, speech synthesis and recognition, intelligent web search, financial forecasting, personal healthcare, traffic navigation, and many other consumer applications. Key enablers that have propelled ML to the forefront of AI research include availability of vast volumes of data, algorithmic advancements that have enabled effective training of deep neural networks, and accessibility and affordability of powerful computing resources. Consequently, novel learning paradigms have been developed beyond the classical discriminative supervised, unsupervised, and semisupervised approaches. Notable novel learning paradigms include reinforcement learning, transfer learning, lifelong learning, generative adversarial learning, and more.