Explaining Groups of Points in Low-Dimensional Representations

Authors: Gregory Plumb, Jonathan Terhorst, Sriram Sankararaman, Ameet Talwalkar

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

Reproducibility Variable Result LLM Response
Research Type Experimental Empirically, we demonstrate that TGT is able to identify explanations that accurately explain the model while being relatively sparse, and that these explanations match real patterns in the data. We demonstrate the usefulness of TGT with a series of experiments on synthetic, UCI, and single-cell RNA datasets. In our experiments, we measure the effectiveness of explanations using correctness and coverage, with sparsity as a proxy metric for interpretability, and we compare the patterns the explanations find to those we expect to be in the data.
Researcher Affiliation Collaboration Gregory Plumb 1 Jonathan Terhorst 2 Sriram Sankararaman 3 Ameet Talwalkar 1 4 1Carnegie Mellon University 2University of Michigan 3University of California, Los Angeles 4Determined AI.
Pseudocode Yes Algorithm 1 TGT: Calculating GCEs with a Reference Group. Algorithm 2 How to construct any explanation between an arbitrary pair of groups, ti!j... Algorithm 3 How to update the basis explanations...
Open Source Code Yes Code for all algorithms and experiments is available at https://github.com/GDPlumb/ELDR
Open Datasets Yes We use the model from (Ding et al., 2018) on the UCI Iris, Boston Housing, and Heart Disease datasets (Dua & Graff, 2017) and a single-cell RNA dataset (Shekhar et al., 2016)
Dataset Splits No The paper describes using datasets and training models but does not provide specific details on train, validation, or test splits (e.g., percentages or sample counts).
Hardware Specification No The paper does not provide specific hardware details (e.g., CPU, GPU models, or memory specifications) used for running the experiments.
Software Dependencies No The paper mentions using an 'autodifferentiation software' but does not provide specific software names with version numbers for reproducibility.
Experiment Setup Yes Algorithm 1 TGT... Input: Model: r Group Means: xi (feature space) and ri (representation space)... l1 Regularization Weight: λ Learning Rate: Initialize: δ1, . . . , δl 1 to vectors of 0...