The traditional diagnosis of neuropsychiatric disorders mainly depends on the subjective evaluation of specialists,neuropsychological test,biochemical examination and other methods,which lacks objective,accurate and intelligent biomarkers.With the rapid development of neuroimaging and artificial intelligence technology,unsupervised learning has been widely used in the auxiliary diagnosis of neuropsychiatric disorders for it has the advantages of independence of external labels,high model generalization,and automatic feature extraction.Compared with the traditional supervised learning methods,unsupervised learning is more capable of achieving objective,accurate and intelligent diagnosis of neuropsychiatric disorders.Herein an overview on the applications of unsupervised learning in the auxiliary diagnosis of neuropsychiatric disorders is provided,summarizing the findings of unsupervised learning in Alzheimer's disease,schizophrenia,major depressive disorder,and autism spectrum disorder,and discussing the research challenges such as insufficient image processing capability,small sample size,insufficient biochemical index data.The corporation with neural network,multi-site large sample size,and deep fusion of multidimensional data are the development trends of unsupervised learning method.