Book excerpt: Nowadays, exploring the brain-behavior relationship via MRI, EEG, fNIRS, and MEG has become a research hotspot further accelerated by the emergence of large-sample open-source datasets, such as UK Biobank, Human Connectome Project, the Adolescent Brain Cognitive Development, the National Institute of Mental Health (NIMH) Intramural Healthy Volunteer Dataset, the TUH EEG CORPUS, and many other multimodal datasets. Many prior studies have conducted various prediction tasks in different populations (from infants to adults; from healthy subjects to patients) with miscellaneous imaging modalities, however, to construct a precise, generalizable, and reproducible brain-behavior relationship is still facing many challenges, for example, individual variability, multi-site heterogeneity, imaging result interpretability, model generalization, low prediction performance, and lack of clinical applications