Exact and Approximate Inference for Annotating Graphs with Structural SVMs

Training processes of structured prediction models such as structural SVMs involve frequent computations of the maximum-a-posteriori (MAP) prediction given a parameterized model. For specific output s