REVIEW 6 cited by
Learning Important Features Through Propagating Activation Differences
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
read the original abstract
The purported "black box" nature of neural networks is a barrier to adoption in applications where interpretability is essential. Here we present DeepLIFT (Deep Learning Important FeaTures), a method for decomposing the output prediction of a neural network on a specific input by backpropagating the contributions of all neurons in the network to every feature of the input. DeepLIFT compares the activation of each neuron to its 'reference activation' and assigns contribution scores according to the difference. By optionally giving separate consideration to positive and negative contributions, DeepLIFT can also reveal dependencies which are missed by other approaches. Scores can be computed efficiently in a single backward pass. We apply DeepLIFT to models trained on MNIST and simulated genomic data, and show significant advantages over gradient-based methods. Video tutorial: http://goo.gl/qKb7pL, ICML slides: bit.ly/deeplifticmlslides, ICML talk: https://vimeo.com/238275076, code: http://goo.gl/RM8jvH.
Forward citations
Cited by 6 Pith papers
-
Interpreting Parton Distributions with Shapley Values
Exact Shapley values on PDF flavors treat χ² as the cooperative payoff, revealing data constraints and an unexpected intermediate-x gluon insensitivity.
-
Towards Verified and Targeted Explanations through Formal Methods
ViTaX certifies targeted semifactual robustness: a minimal feature subset can be perturbed by ε without flipping a neural network from class y to a user-specified high-risk class t.
-
Structure-Aware Compound-Protein Affinity Prediction via Graph Neural Networks with Group Lasso Regularization
Adding common-plus-uncommon substructure masks and group lasso or sparse group lasso losses to activity-cliff GNNs improves per-target pIC50 prediction and attribution consistency on six tyrosine kinases.
-
inMOTIFin: a lightweight end-to-end simulation software for regulatory sequences
inMOTIFin is a modular Python package for simulating regulatory DNA sequences with controlled motif grammar, co-occurrence, positions, orientations, and direct sequence edits.
-
Towards a Science of Causal Interpretability in Deep Learning for Software Engineering
The dissertation presents docode, a causal interpretability method for neural code models, and uses a case study to show that some correlations between code properties and model performance are confounded rather than causal.
-
ShaTS: A Shapley-based Explainability Method for Time Series Artificial Intelligence Models applied to Anomaly Detection in Industrial Internet of Things
ShaTS computes Shapley attributions directly on semantic groups of time-series features, improving sensor- and process-level anomaly explanations over post hoc SHAP on the SWaT dataset.
Discussion (0). Continue with ORCID to comment.