The paper introduces a simple, effective, and efficient method for unsupervised domain adaptation called CORrelation ALignment (CORAL). CORAL minimizes domain shift by aligning the second-order statistics of source and target distributions without requiring any target labels. The method is straightforward to implement, requiring only the computation of covariance statistics and applying a linear transformation to the source features. Extensive evaluations on standard benchmark datasets show that CORAL outperforms many existing methods, including complex deep neural network approaches, in tasks such as object recognition and sentiment prediction. The authors demonstrate that CORAL works well with both shallow and deep features, and can be easily integrated into deep architectures. The method is particularly valuable in the era of deep learning, where high-level deep features are more strongly correlated and thus benefit more from CORAL's alignment technique.The paper introduces a simple, effective, and efficient method for unsupervised domain adaptation called CORrelation ALignment (CORAL). CORAL minimizes domain shift by aligning the second-order statistics of source and target distributions without requiring any target labels. The method is straightforward to implement, requiring only the computation of covariance statistics and applying a linear transformation to the source features. Extensive evaluations on standard benchmark datasets show that CORAL outperforms many existing methods, including complex deep neural network approaches, in tasks such as object recognition and sentiment prediction. The authors demonstrate that CORAL works well with both shallow and deep features, and can be easily integrated into deep architectures. The method is particularly valuable in the era of deep learning, where high-level deep features are more strongly correlated and thus benefit more from CORAL's alignment technique.