Incremental learning from chunk data for IDR/QR
作者：
Highlights：
• We propose a new implementation of the batch IDR/QR method.
• Our new implementation is theoretically equivalent to the original one but is more efficient.
• Based on our new implementation of batch IDR/QR, we propose the chunk IDR method.
• Chunk IDR is capable of processing multiple data instances at a time.
• Chunk IDR can accurately update the discriminant vectors when new data items are added dynamically.
摘要
•We propose a new implementation of the batch IDR/QR method.•Our new implementation is theoretically equivalent to the original one but is more efficient.•Based on our new implementation of batch IDR/QR, we propose the chunk IDR method.•Chunk IDR is capable of processing multiple data instances at a time.•Chunk IDR can accurately update the discriminant vectors when new data items are added dynamically.
论文关键词：Feature extraction,Dimensionality reduction,Linear discriminant analysis,Small sample size problem,Incremental learning
论文评审过程：Received 25 September 2013, Revised 15 September 2014, Accepted 5 January 2015, Available online 19 February 2015.
论文官网地址：https://doi.org/10.1016/j.imavis.2015.01.002
