Dictionary Learning for Neural Network Interpretability

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

• Neural Network Interpretabilityconcept0.864
  The field aimed at understanding what neural networks have learned; characterized as pre-paradigmatic in this paper
• Sparse interpretability in neural networksconcept0.797
  VPD achieves sparse, interpretable parameter subcomponents with improved sparsity-reconstruction tradeoff.
• Neural Decoding Programframework0.776
• Neural Decodingframework0.772
• Sparse Dictionary Learningmethod0.769
  General method for finding overcomplete sparse decompositions; the paper uses sparse autoencoders as an approximation
• Sparse Autoencoder for Dictionary Learningframework0.761
  Primary method introduced: trains a one-hidden-layer MLP with L1 sparsity penalty to decompose model activations into overcomplete feature dictionaries
• Neural Networksconcept0.759
• Learned features reflect the functionality of the model and not just the data distribution, as evidenced by interpretable downstream effects not used in dictionary learningclaim0.758
  Authors argue features are model properties because logit effects and ablations are consistent with feature interpretations