This chapter provides a comprehensive exploration of performance metrics for various machine learning algorithms, including supervised, unsupervised, generative AI, NLP, and reinforcement learning. It delves into the selection of appropriate metrics and addresses critical issues such as overfitting, underfitting, outlier handling, confusion metrics, and monotonicity. The chapter also covers best practices for dataset splitting, cross-validation techniques, and balancing bias and variance. Recent developments in RAG for LLMs and the Hallucination Index Report for foundation models are discussed. An in-depth examination of training methodologies for different machine learning models is presented, along with hyperparameter tuning strategies. The chapter explores techniques such as early stopping and regularization to enhance model generalizability, as well as model pruning and optimization of computational cost.