AI Paper Atlas

Research map / Learning theory and algorithms / Combinatorial and search optimization

Sufficient Reasons for Classifier Decisions in the Presence of Domain Constraints

Niku Gorji, Sasha Rubin

AAAI 2022 · Learning theory and algorithms › Combinatorial and search optimization · approach: Structure and complexity of constraint reasoning · 29 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Treat constraint-violating instances as undefined rather than ignoring the constraintBoolean classifier sufficient reasons · Constrained decision functionsreason succinctnessImproved
Represent constraints using implication (κ → φ), rather than ignoring themBoolean classifier sufficient reasons · Constrained decision functionsreason parsimonyImproved
Represent constraints using conjunction (κ ∧ φ), rather than implicationBoolean classifier sufficient reasons · Constrained decision functionsreason parsimonyHurt
Use one-hot constraints to simplify explanationsCRASH trial RPART decision tree · CRASH trial, selected positive instancereason sizeImproved
Use the no-overlap constraint for X and O piecesTic-Tac-Toe RPART decision tree · Tic-Tac-Toe, selected positive instancenumber of reasons: 8 -> 11Improved
Use valid-play constraints, including move order and alternationTic-Tac-Toe RPART decision tree · Tic-Tac-Toe, selected positive instancenumber of reasons: 46Mixed

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy