Export Reviews, Discussions, Author Feedback and Meta-Reviews

"NIPS Neural Information Processing Systems 8-11th December 2014, Montreal, Canada",,, "Paper ID:","1663" "Title:","Learning Distributed Representations for Structured Output Prediction" Current Reviews First provide a summary of the paper, and then address the following criteria: Quality, clarity, originality and significance. The paper investigates finding an encoding of labels as real valued/low dimensional vectors (as opposed to indicator vectors) in Structured Output (SO) prediction setting. Tensor products are used for non-unary potentials. The label representations are learned jointly with weight vectors using an alternating optimization procedure, minimizing the trace norm of the label representation matrix and minimizing the hinge loss on the data wrt feature weights. Experiments on multiclass classification and sequence labeling are provided.