Science Advances

Supplementary Materials

The PDF file includes:

  • Supplementary Text
  • Fig. S1. Comparison of different CNN architectures.
  • Fig. S2. Visual features associated with different emotion schemas.
  • Fig. S3. Dimensionality of CNN predictions in the holdout dataset estimated with PCA and clustering of classification errors in 20-way classification.
  • Fig. S4. Surface renders depict where decreases (blue) or increases (red) in fMRI activation are predictive of activation in emotion category units of EmoNet.
  • Fig. S5. Information about emotion schema is distributed across human visual cortex.
  • Fig. S6. Decoding EmoNet activation using fMRI responses from different visual areas and a model comprising the entire occipital lobe.
  • Fig. S7. Results of simulations using repeated random subsampling to assess sample size and power for fMRI experiment I.
  • Fig. S8. Classification of images containing dogs from ImageNet (68).
  • Fig. S9. Simulated experiments used to evaluate the bias of the discriminable cluster identification method.
  • Legends for movies S1 to S3
  • References (6168)

Download PDF

Other Supplementary Material for this manuscript includes the following:

  • Movie S1 (.mp4 format). Model predictions for action trailers.
  • Movie S2 (.mp4 format). Model predictions for horror trailers.
  • Movie S3 (.mp4 format). Model predictions for romantic comedy trailers.

Files in this Data Supplement: