Processing math: 100%

Discrete Laplacian Distribution

Defined over all integers for a>0

p(k)=tanh(a2)ea|k|,F(x)={ea(x+1)ea+1x<0,1eaxea+1x0.G(q)={1alog[q(ea+1)]1q<11+ea,1alog[(1q)(1+ea)]q11+ea.
M(t)=tanh(a2)k=etkea|k|=C(1+k=1e(t+a)k+1e(ta)k)=tanh(a2)(1+e(t+a)1e(t+a)+eta1eta)=tanh(a2)sinhacoshacosht.

Thus,

μn=M(n)(0)=[1+(1)n]Lin(ea)

where Lin(z) is the polylogarithm function of order n evaluated at z.

h[X]=log(tanh(a2))+asinha

Implementation: scipy.stats.dlaplace