neural additive model (NAM) [15]. The NAM is a differentiable (gradient-based) additive model which can adapt as the estimated latent treatment benefits are updated during training.

S-learner strategy

GAMs [12]

the inner product 〈g(X1,X2),X2〉<math><mrow is="true"><mo stretchy="false" is="true">〈</mo><mi is="true">g</mi><mo stretchy="false" is="true">(</mo><msub is="true"><mrow is="true"><mi is="true">X</mi></mrow><mrow is="true"><mn is="true">1</mn></mrow></msub><mo is="true">,</mo><msub is="true"><mrow is="true"><mi is="true">X</mi></mrow><mrow is="true"><mn is="true">2</mn></mrow></msub><mo stretchy="false" is="true">)</mo><mo is="true">,</mo><msub is="true"><mrow is="true"><mi is="true">X</mi></mrow><mrow is="true"><mn is="true">2</mn></mrow></msub><mo stretchy="false" is="true">〉</mo></mrow></math> converts this potential benefit into an estimated scalar benefit under the observed treatments X2<math><mrow is="true"><msub is="true"><mrow is="true"><mi is="true">X</mi></mrow><mrow is="true"><mn is="true">2</mn></mrow></msub></mrow></math>.