@article{b6490d6946c24d3e913f84a8b361d394,
title = "Sensitivity of Sparse Codes to Image Distortions",
abstract = "Sparse coding has been proposed as a theory of visual cortex and as an unsupervised algorithm for learning representations. We show empirically with the MNIST data set that sparse codes can be very sensitive to image distortions, a behavior that may hinder invariant object recognition. A locally linear analysis suggests that the sensitivity is due to the existence of linear combinations of active dictionary elements with high cancellation. A nearest-neighbor classifier is shown to perform worse on sparse codes than original images. For a linear classifier with a sufficiently large number of labeled examples, sparse codes are shown to yield higher accuracy than original images, but no higher than a representation computed by a random feedforward net. Sensitivity to distortions seems to be a basic property of sparse codes, and one should be aware of this property when applying sparse codes to invariant object recognition.",
author = "Kyle Luther and Seung, {H. Sebastian}",
note = "Funding Information: We thank L. and J. Jackel, Lawrence Saul, and Runzhe Yang for their helpful insights and discussions. We also thank Dylan Paiton and Bruno Olshausen for their feedback that helped to make this letter{\textquoteright}s claims more precise and the experimental analysis more rigorous. This research was supported by the Intelligence Advanced Research Projects Activity (IARPA) via the Department of Interior/Interior Business Center (DoI/IBC) contract numbers D16PC0005, NIH/NIMH RF1MH117815, and RF1MH123400. The U.S. government is authorized to reproduce and distribute reprints for governmental purposes notwithstanding any copyright annotation thereon. The views and conclusions contained here are our own and should not be interpreted as necessarily representing the official policies or endorsements, either expressed or implied, of IARPA, DoI/IBC, or the U.S. government. Funding Information: We thank L. and J. Jackel, Lawrence Saul, and Runzhe Yang for their helpful insights and discussions. We also thank Dylan Paiton and Bruno Olshausen for their feedback that helped to make this letter{\textquoteright}s claims more precise and the experimental analysis more rigorous. This research was supported by the Intelligence Advanced Research Projects Activity (IARPA) via the Department of Interior/Interior Business Center (DoI/IBC) contract numbers D16PC0005, NIH/NIMHRF1MH117815, and RF1MH123400. The U.S. government is authorized to reproduce and distribute reprints for governmental purposes notwithstanding any copyright annotation thereon. The views and conclusions contained here are our own and should not be interpreted as necessarily representing the official policies or endorsements, either expressed or implied, of IARPA, DoI/IBC, or the U.S. government. Publisher Copyright: {\textcopyright} 2022 Massachusetts Institute of Technology.",
year = "2022",
month = jun,
day = "16",
doi = "10.1162/neco_a_01513",
language = "English (US)",
volume = "34",
pages = "1616--1635",
journal = "Neural Computation",
issn = "0899-7667",
publisher = "MIT Press Journals",
number = "7",
}