Rejoinder: New Objectives for Policy Learning.

One example is of course Graph HT <math display="inline" xmlns="http://www.w3.org/1998/Math/MathML"><mrow><mi>w</mi><mo>=</mo><mn mathvariant="bold">1</mn></mrow></math> ht . I want to first reemphasize that there is nothing special about the original population (Graph HT <math display="inline" xmlns="http://www.w3.org/1998/Math/MathML"><mrow><mi>w</mi><mo>=</mo><mn mathvariant="bold">1</mn></mrow></math> ht ) with regard to curvature: it may better, it may be worse. There appears to be no method that adapts to each setting except for the oracle Graph HT <math display="inline" xmlns="http://www.w3.org/1998/Math/MathML"><mrow><msup><mrow><mi>w</mi></mrow><mo>†</mo></msup></mrow></math> ht method, which is of course not implementable. The flat-value-surface example that LLL pose at the end of their Section 2.2 can happen for I w i = I w i SB 0 sb just as much as it can happen for Graph HT <math display="inline" xmlns="http://www.w3.org/1998/Math/MathML"><mrow><mi>w</mi><mo>=</mo><mn mathvariant="bold">1</mn></mrow></math> ht . [Extracted from the article]