This repository contains the code for our proposed method "Multi-source Feature Alignment and Label Rectification" (MFA-LR), which has been published in the paper "Learning a Robust Unified Domain Adaptation Framework for Cross-subject EEG-based Emotion Recognition."