The Shapley Taylor Interaction Index

Authors: Mukund Sundararajan, Kedar Dhamdhere, Ashish Agarwal

ICML 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Though our key contributions and evaluations are mainly theoretical, we demonstrate the applicability of our work in Section 5, which studies models for three tasks (sentiment analysis, random forest regression, and question answering). We identify certain interesting interactions.
Researcher Affiliation Industry Kedar Dhamdhere 1 Ashish Agarwal 1 Mukund Sundararajan 1 1Google LLC.
Pseudocode No The paper does not contain structured pseudocode or algorithm blocks. The methods are described through mathematical equations and textual explanations.
Open Source Code No The paper does not provide any explicit statement about releasing source code or a link to a code repository for the methodology described.
Open Datasets Yes The second is a random forest regression model built to predict house prices. We use the Boston house price dataset ((10)). [...] The third model, called QANet (32), solves reading comprehension, i.e., identifying a span from a context paragraph as an answer to a question; it uses the SQu AD dataset ((20)).
Dataset Splits No The paper states: "There are 506 data points. We split the data into 385 training and 121 test examples." This specifies training and test splits but does not explicitly mention a validation split.
Hardware Specification No The paper does not provide specific hardware details (e.g., GPU/CPU models, memory) used for running its experiments. It only mentions general models and tasks.
Software Dependencies No The paper states: "We used scikit-learn to train a random forest model." However, it does not specify any version numbers for scikit-learn or any other software dependencies.
Experiment Setup Yes To ablate a word, we zero out the embedding for that word. [...] When we ablate a feature, we replace it by its training-data mean.