A PAC-Bayesian Perspective on Structured Prediction with Implicit Loss Embeddings
This work provides theoretical insights and new learning algorithms for researchers working on structured prediction problems, particularly within the ILE framework.
This paper proposes a novel PAC-Bayesian perspective on the Implicit Loss Embedding (ILE) framework for structured prediction, deriving two generalization bounds for risk and excess risk. These bounds are then used to derive two learning algorithms, which are implemented and analyzed.
Many practical machine learning tasks can be framed as Structured prediction problems, where several output variables are predicted and considered interdependent. Recent theoretical advances in structured prediction have focused on obtaining fast rates convergence guarantees, especially in the Implicit Loss Embedding (ILE) framework. PAC-Bayes has gained interest recently for its capacity of producing tight risk bounds for predictor distributions. This work proposes a novel PAC-Bayes perspective on the ILE Structured prediction framework. We present two generalization bounds, on the risk and excess risk, which yield insights into the behavior of ILE predictors. Two learning algorithms are derived from these bounds. The algorithms are implemented and their behavior analyzed, with source code available at \url{https://github.com/theophilec/PAC-Bayes-ILE-Structured-Prediction}.