Pith. sign in

REVIEW

On the Veracity of Local, Model-agnostic Explanations in Audio Classification: Targeted Investigations with Adversarial Examples

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

arxiv 2107.09045 v2 pith:PBPFIGGY submitted 2021-07-19 eess.AS cs.AIcs.SD

On the Veracity of Local, Model-agnostic Explanations in Audio Classification: Targeted Investigations with Adversarial Examples

classification eess.AS cs.AIcs.SD
keywords explanationslimeclassificationexplanationlocalmodelpredictionwhat
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Local explanation methods such as LIME have become popular in MIR as tools for generating post-hoc, model-agnostic explanations of a model's classification decisions. The basic idea is to identify a small set of human-understandable features of the classified example that are most influential on the classifier's prediction. These are then presented as an explanation. Evaluation of such explanations in publications often resorts to accepting what matches the expectation of a human without actually being able to verify if what the explanation shows is what really caused the model's prediction. This paper reports on targeted investigations where we try to get more insight into the actual veracity of LIME's explanations in an audio classification task. We deliberately design adversarial examples for the classifier, in a way that gives us knowledge about which parts of the input are potentially responsible for the model's (wrong) prediction. Asking LIME to explain the predictions for these adversaries permits us to study whether local explanations do indeed detect these regions of interest. We also look at whether LIME is more successful in finding perturbations that are more prominent and easily noticeable for a human. Our results suggest that LIME does not necessarily manage to identify the most relevant input features and hence it remains unclear whether explanations are useful or even misleading.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.