Equivalence of logarithm-convex.

52 Views Asked by At

‎we have: A function ‎$‎f:I‎‎\rightarrow‎‎\mathbb{R}‎^+‎$ ‎is ‎log-convex ‎if ‎and ‎only ‎if‎ ‎\begin{align*}‎‎ ‎f(‎\lambda‎ x + u y)\leq f^{‎\lambda‎}(x) f^u (y) ‎\end{align*}‎‎ ‎for ‎$‎x, y\in I‎$ ‎and ‎‎$‎‎\lambda‎, u>0‎$ ‎with ‎‎$‎‎\lambda +‎ u‎ =‎ ‎1‎$‎.‎

‎Now, my question is:‎‎ why? The reformulation of log-convexity implied by above inequality is equivalent to the following working definition: the function ‎$‎f:I‎‎\rightarrow‎‎\mathbb{R}‎$ ‎is ‎log-convex ‎if ‎and ‎only ‎if ‎for ‎all ‎‎$‎x, y, z\in I‎$ ‎with ‎‎$‎x<y<z‎$‎‎ ‎\begin{align*}‎‎ ‎f^{z-x}(y)\leq f^{z-y}(x) f^{y-x}(z). ‎\end{align*}

1

There are 1 best solutions below

3
On BEST ANSWER

The two are equivalent. If $x <y<z$ we can write $y=\lambda z+(1-\lambda)x$ where $\lambda=\frac {y-x} {z-x}$. Applying the first definition to this we get the second one. [ We get $f(y) \leq f(z)^{\frac {y-x} {z-x}} f(x)^{(1-\frac {y-x} {z-x})}$. Rise to power $z-x$]. Conversely, give, $x<z$ and $\lambda ,u >0$ with $\lambda +u =1$ take $y=\lambda z+(1-\lambda)x$. The $x <y<z$ and we can use the second definition to get the first (with $z$ in place of $y$).