Building recombining trinomial trees for time-homogeneous diffusion processes (Q2279897)

Here is considered a diffusion \[X_0=x_0\in\mathbb{R}dX_t= f(X_t)dt+g(X_t)dW_t,\tag{1}\] where \(W\) is a Brownian motion, \(f\) and \(g\) are functions such that (1) admits a unique strong solution. After a short introduction giving some bibliographic elements, a new method for pricing a payoff \(h,(X_T)\), where \(T\) is the maturity, is proposed. Actually, a recombining trinomial tree allows to approximate (1) in distribution. Since the diffusion is time homogeneous, only the transition probabilities at two nodes at each time are needed. Some assumptions are made and under these assumptions a converging algorithm is established. A terminal section proposes some numerical examples, for the diffusion defined with \(f(x)=\kappa(\theta-x)\), \(g(x)=\sigma\ln(1+X_t)\).