Maybe I'm just being dull, but it seems there's an error (or at least a discrepancy) in the proof that the maximum entropy continuous distribution defined on the whole real line with a given variance is the Gaussian, which appears in the Overthinking box on pp. 314-315.
The definitions of Dₖₗ(q, p) and H(q, p) in this proof seem to reverse the signs of the ones given earlier in the book (see, e.g., the Overthinking box on p. 212). Because both are reversed, the proof still goes through, but it's harder to follow than it might otherwise be.
Maybe I'm just being dull, but it seems there's an error (or at least a discrepancy) in the proof that the maximum entropy continuous distribution defined on the whole real line with a given variance is the Gaussian, which appears in the Overthinking box on pp. 314-315.
The definitions of Dₖₗ(q, p) and H(q, p) in this proof seem to reverse the signs of the ones given earlier in the book (see, e.g., the Overthinking box on p. 212). Because both are reversed, the proof still goes through, but it's harder to follow than it might otherwise be.