This chapter presents neural network approaches for inverse polymer design, enabling the generation of polymers with or without specific constraints. It introduces unconstrained generation, which explores potential polymer structures by sampling from latent spaces, and constrained generation, where specific property or structural requirements guide the process. The chapter discusses diverse datasets, evaluation metrics, and neural network architectures such as Generative Adversarial Networks (GANs), Variational Autoencoders (VAEs), and Diffusion Models, adapted for sequence- and graph-based polymer representations. Advanced techniques for incorporating constraints–such as property vectors, polymer substructures, or natural language descriptions–are detailed through adaptive normalization and cross-attention mechanisms.