1: \begin{abstract}

2: 

3: In \cite{NeriSchneider2009} we presented a method to recover the Maximum Entropy Density (MED) inferred from prices of call and digital options on a set of $n$ strikes.

4: To find the MED we need to numerically invert a one-dimensional function for $n$ values and a Newton-Raphson method is suggested.

5: In this note we revisit this inversion problem and show that it can be rewritten in terms of the Langevin function for which numerical approximations of its inverse are known.

6: 

7: The approach of \cite{NeriSchneider2009} is very similar to that of Buchen and Kelly (BK) \cite{BuchenKelly1996} with the difference that BK only requires call option prices.

8: Then, in \cite{NeriSchneider2011} we presented another approach which uses call prices only and recovers the same density as BK with a few advantages, notably, numerical stability.

9: \cite{NeriSchneider2011} provides a detailed analysis of convergence and, in particular, gives various estimates of how far (in different senses) the iterative algorithm is from the solution.

10: These estimates rely on a constant $m>0$.

11: The larger $m$ is the better the estimates will be.

12: A concrete value of $m$ is suggested in \cite{NeriSchneider2011}, and this note provides a sharper value.

13: 

14: \bigskip

15: 

16: \keywords{Entropy \and Information Theory \and $I$-Divergence \and Asset Distribution \and Option Pricing}

17: 

18: \bigskip

19: 

20: % MSC2000

21: \subclass{91B24 \and 91B28 \and 91B70 \and 94A17}

22: %http://www.ams.org/msc/

23: %91B24 Price theory and market structure

24: %91B28 Finance, portfolios, investment

25: %91B70 Stochastic models

26: %94A17 Measures of information, entropy

27: 

28: \bigskip

29: 

30: % JEL

31: \JEL{C16 \and C63 \and G13}

32: %C16 Specific Distributions

33: %C63 Computational Techniques; Simulation Modeling

34: %G13 Contingent Pricing; Futures Pricing

35: \end{abstract}