The generalization ability of Convolutional neural networks (CNNs) for biometrics drops greatly due to the adverse ef- fects of various occlusions. To this end, we propose a novel unified framework integrated the merits of both CNNs and graphical models to learn dynamic graph representations for occlusion problems in biometrics, called Dynamic Graph Representation (DGR). Convolutional features onto certain regions are re-crafted by a graph generator to establish the connections among the spatial parts of biometrics and build Feature Graphs based on these node representations. Each node of Feature Graphs corresponds to a specific part of the input image and the edges express the spatial relationships be- tween parts. By analyzing the similarities between the nodes, the framework is able to adaptively remove the nodes rep- resenting the occluded parts. During dynamic graph match- ing, we propose a novel strategy to measure the distances of both nodes and adjacent matrixes. In this way, the proposed method is more convincing than CNNs-based methods be- cause the dynamic graph method implies a more illustrative and reasonable inference of the biometrics decision. Experi- ments conducted on iris and face demonstrate the superior- ity of the proposed framework, which boosts the accuracy of occluded biometrics recognition by a large margin com- paring with baseline methods. The code is available at https: //github.com/RenMin1991/Dyamic Graph Representation