Unsupervised learning is second one of the three major paradigms of machine learning. We first give the definition of unsupervised learning, explain its working principle using formal and illustrated descriptions, and introduce the classic tasks of unsupervised learning. Due to the outstanding ability of humans in unsupervised learning, generative models beyond the classic unsupervised learning have become the important topics. In this regard, we focus on the introduction to the generative models proposed in recent years, namely energy models, autoencoding models, generative adversarial models, normalizing flow models, autoregressive models, and diffusion models. It can be thought that the generative models have given birth to large language models.