training-and-inference
Training adjusts model parameters to fit known data, while inference applies the trained model to new, unseen examples to generate predictions.
1 chapter across 1 book
Deep Learning (MIT Press Essential Knowledge)John D. Kelleher
The chapter introduces deep learning as a powerful subfield of artificial intelligence that uses large neural networks to model complex functions from data, highlighting its transformative impact exemplified by AlphaGo's success. It explains foundational machine learning concepts including training and inference, inductive bias, and the challenges of underfitting and overfitting, emphasizing the need for large datasets and appropriate model assumptions. The chapter also outlines key machine learning paradigms—supervised, unsupervised, and reinforcement learning—and positions deep learning's advantage in automatic feature learning, concluding with an overview of future research directions and challenges such as model interpretability and achieving general intelligence.