Beyond one-hot encoding: Lower dimensional target embedding

作者：

Highlights：

• We propose Error-Correcting Ouput Codes as an alternative to one-hot for memory-constrained deep learning.

• The proposed approach relies on an eigenrepresentation of the class manifold.

• Networks trained with our approach converge faster than using one-hot on multiple datasets.

摘要

•We propose Error-Correcting Ouput Codes as an alternative to one-hot for memory-constrained deep learning.•The proposed approach relies on an eigenrepresentation of the class manifold.•Networks trained with our approach converge faster than using one-hot on multiple datasets.

论文关键词：Error correcting output codes,Output embeddings,Deep learning,Computer vision

论文评审过程：Received 23 May 2017, Accepted 27 April 2018, Available online 11 May 2018, Version of Record 30 May 2018.

论文官网地址：https://doi.org/10.1016/j.imavis.2018.04.004