Machine Learning for Logic Synthesis

In this chapter we will cover several applications of machine learning in logic synthesis algorithms for electronic design automation (EDA). We will discuss how machine learning models can learn to guide optimization algorithms or learn optimization policies directly. We will study supervised learning or reinforcement learning formulations for various logic synthesis algorithms. We will discuss the architecture of corresponding machine learning models. For reinforcement learning formulations, we will discuss state-action spaces and their representation. We will also discuss scalability considerations for reinforcement learning applications in logic synthesis.