Figures
Priority-based transformation of stimulus representations in hidden layer of an RNN
PCA visualization of LSTM hidden layer activity of an example 7-hidden-unit RNN while performing a 2-back working memory task. Visualized are the representations of different input stimuli (each represented with a different color) when in an unprioritized state in working memory. Solid gray arrow denotes the stimulus coding axis, which will rotate counterclockwise in the image plane (defined by the 1st and 2nd PCs) as the stimulus transitions to a prioritized state (indicated by the dashed gray arrow).
Image Credit: Quan Wan, Po'stl'ab, University of Wisconsin-Madison
Citation: (2022) PLoS Computational Biology Issue Image | Vol. 18(6) June 2022. PLoS Comput Biol 18(6): ev18.i06. https://doi.org/10.1371/image.pcbi.v18.i06
Published: June 30, 2022
Copyright: © 2022 . This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
PCA visualization of LSTM hidden layer activity of an example 7-hidden-unit RNN while performing a 2-back working memory task. Visualized are the representations of different input stimuli (each represented with a different color) when in an unprioritized state in working memory. Solid gray arrow denotes the stimulus coding axis, which will rotate counterclockwise in the image plane (defined by the 1st and 2nd PCs) as the stimulus transitions to a prioritized state (indicated by the dashed gray arrow).
Image Credit: Quan Wan, Po'stl'ab, University of Wisconsin-Madison