notesum.ai

Published at November 10

BayesNAM: Leveraging Inconsistency for Reliable Explanations

cs.LG

cs.AI

cs.NE

Released Date: November 10, 2024

Authors: Hoki Kim, Jinseong Park, Yujin Choi, Seungyun Lee, Jaewook Lee

Arxiv: http://arxiv.org/abs/2411.06367v1

Refer to caption

Model

COMPAS

(AUC

\uparrow

)

Credit

(AUC

\uparrow

)

FICO

(AUC

\uparrow

)

Boston

(RMSE

\downarrow

)

CA Housing

(RMSE

\downarrow

)

Log./Lin. Reg.

0.699

\pm

0.005

0.977

\pm

0.004

0.706

\pm

0.005

5.517

\pm

0.009

0.731

\pm

0.010

CART

0.776

\pm

0.005

0.956

\pm

0.005

0.784

\pm

0.002

4.133

\pm

0.004

0.712

\pm

0.007

XGBoost

0.743

\pm

0.012

0.980

\pm

0.005

0.795

\pm

0.001

3.155

\pm

0.009

0.531

\pm

0.011

EBM

0.764

\pm

0.009

0.978

\pm

0.007

0.793

\pm

0.005

3.301

\pm

0.005

0.558

\pm

0.012

NAM

0.769

\pm

0.011

0.989

\pm

0.007

0.804

\pm

0.003

3.567

\pm

0.012

0.556

\pm

0.009

NAM+Ens

0.771

\pm

0.005

0.990

\pm

0.004

0.804

\pm

0.002

3.555

\pm

0.006

0.554

\pm

0.003

BayesNAM

0.784

\pm

0.009

0.991

\pm

0.003

0.804

\pm

0.001

3.620

\pm

0.011

0.556

\pm

0.007