Ito 1
如果F是关于随机变量X的函数(其中t是隐式呈现的),比如 F = X 2 , F = e X F=X^2,F=e^X F=X2,F=eX等等,则根据Ito 1有随机微分:
d F = d F d X d X + 1 2 d 2 F d X 2 d t dF=\frac{dF}{dX}dX+\frac{1}{2}\frac{d^2F}{dX^2}dt dF=dXdFdX+21dX2d2Fdt
注意:
- d F d X \frac{dF}{dX} dXdF和 d 2 F d X 2 \frac{d^2F}{dX^2} dX2d2F都是通过普通微分求导的方法得到,代入式中即可。其余的Ito也是一样。
- 不需要考虑 d X d t \frac{dX}{dt} dtdX(不适用链式法则),是因为X处处连续,处处不可导。 d X d t \frac{dX}{dt} dtdX并不存在。
Ito 2
如果F是关于随机变量X的函数(其中t是显式呈现的),比如 F = t 2 + X t 2 F=t^2+X_t^2 F=t2+Xt2等,则需要进行二维的泰勒展开:
令 t → t + d t , X t → X t + d X t t\rightarrow t+dt,X_t\rightarrow X_t+dX_t t→t+dt,Xt→Xt+dXt
F ( t + d t , X + d X ) = F ( t , X ) + ∂ F ∂ t d t + ∂ F ∂ X d X + 1 2 ∂ 2 F ∂ X 2 d X 2 + . . . F(t+dt,X+dX)=F(t,X)+\frac{\partial F}{\partial t}dt+\frac{\partial F}{\partial X}dX+\frac{1}{2}\frac{\partial^2F}{\partial X^2}dX^2+... F(t+dt,X+dX)=F(t,X)+∂t∂Fdt+∂X∂FdX+21∂X2∂2FdX2+...
⇒ d F = ∂ F ∂ X d X + ( ∂ F ∂ t + 1 2 ∂ 2 F ∂ X 2 ) d t \Rightarrow dF=\frac{\partial F}{\partial X}dX+(\frac{\partial F}{\partial t}+\frac{1}{2}\frac{\partial^2F}{\partial X^2})dt ⇒dF=∂X∂FdX+(∂t∂F+21∂X2∂2F)dt
伊藤积分(对随机项的积分)
-
从Ito 1开始
∫ 0 t ∂ F ∂ X s d X s = F ( t , X t ) − F ( 0 , X 0 ) − 1 2 ∫ 0 t ∂ 2 F ∂ X s 2 d s \int^t_0\frac{\partial F}{\partial X_s}dX_s=F(t,X_t)-F(0,X_0)-\frac{1}{2}\int_0^t\frac{\partial^2F}{\partial X^2_s}ds ∫0t∂Xs∂FdXs=F(t,Xt)−F(0,X0)−21∫0t∂Xs2∂2Fds
-
从Ito 2开始
∫ 0 t ∂ F ∂ X s d X s = F ( t , X t ) − F ( 0 , X 0 ) − ∫ 0 t ( ∂ F ∂ s + 1 2 ∂ 2 F ∂ X s 2 d s ) \int^t_0\frac{\partial F}{\partial X_s}dX_s=F(t,X_t)-F(0,X_0)-\int_0^t(\frac{\partial F}{\partial s}+\frac{1}{2}\frac{\partial^2F}{\partial X^2_s}ds) ∫0t∂Xs∂FdXs=F(t,Xt)−F(0,X0)−∫0t(∂s∂F+21∂Xs2∂2Fds)
Ito 3
随机变量S是满足几何布朗运动( d S = u S d t + σ S d X dS=uSdt+\sigma SdX dS=uSdt+σSdX),要看S的函数V的随机微分方程是什么样的(时间t是隐式呈现的),则需要利用Ito 3.
根据伊藤乘法表可知: d S 2 = ( a d t + b d X ) 2 = b 2 d t dS^2=(adt+bdX)^2=b^2dt dS2=(adt+bdX)2=b2dt,即 d S 2 = σ 2 S 2 d t dS^2=\sigma^2S^2dt dS2=σ2S2dt
V ( S + d S ) ≈ V ( S ) + d V d S d S + 1 2 d 2 V d S 2 d S 2 V(S+dS)\approx V(S)+\frac{dV}{dS}dS+\frac{1}{2}\frac{d^2V}{dS^2}dS^2 V(S+dS)≈V(S)+dSdVdS+21dS2d2VdS2
将 V ( S ) V(S) V(S)移到等是左边,将 d S dS dS和 d S 2 dS^2 dS2的表达式代入,可得
d V = d V d S ( μ S d t + σ S d X ) + 1 2 d 2 V d S 2 σ 2 S 2 d t dV=\frac{dV}{dS}(\mu Sdt+\sigma SdX)+\frac{1}{2}\frac{d^2V}{dS^2}\sigma^2S^2dt dV=dSdV(μSdt+σSdX)+21dS2d2Vσ2S2dt
= ( μ S d V d S + 1 2 σ 2 S 2 d 2 V d S 2 ) + σ S d V d S d X =(\mu S\frac{dV}{dS}+\frac{1}{2}\sigma^2S^2\frac{d^2V}{dS^2})+\sigma S\frac{dV}{dS}dX =(μSdSdV+21σ2S2dS2d2V)+σSdSdVdX
使用场景:当你遇到一个随机变量的函数,而这个随机变量刚好满足几何布朗运动,则使用Ito 3写出函数的随机微分方程。最常见的服从几何布朗运动的随机变量,就是证券价格S。
Ito 4
随机变量S是满足几何布朗运动( d S = u S d t + σ S d X dS=uSdt+\sigma SdX dS=uSdt+σSdX),要看S的函数V的随机微分方程是什么样的,则需要利用Ito 4。V中的t是显式呈现的(比如: V = t 2 + S 2 ; V = t e S ; V = t 2 + l o g S V=t^2+S^2;V=te^S;V=t^2+logS V=t2+S2;V=teS;V=t2+logS),即: V = V ( t , S ) , S → G B M V=V(t,S),S\rightarrow GBM V=V(t,S),S→GBM
V ( t + d t , S + d S ) = V ( t , S ) + ∂ V ∂ t d t + ∂ V ∂ S d S + 1 2 ∂ 2 V ∂ S 2 d S 2 V(t+dt,S+dS)=V(t,S)+\frac{\partial V}{\partial t}dt+\frac{\partial V}{\partial S}dS+\frac{1}{2}\frac{\partial^2V}{\partial S^2}dS^2 V(t+dt,S+dS)=V(t,S)+∂t∂Vdt+∂S∂VdS+21∂S2∂2VdS2
其中: d S = μ S d t , d S 2 = σ 2 S 2 d t dS=\mu Sdt,dS^2=\sigma^2S^2dt dS=μSdt,dS2=σ2S2dt,代入上式,得到:
d V = ( ∂ V ∂ t + μ S ∂ V ∂ S + 1 2 σ 2 S 2 ∂ 2 V ∂ S 2 ) d t + σ S ∂ V ∂ S d X dV=(\frac{\partial V}{\partial t}+\mu S\frac{\partial V}{\partial S}+\frac{1}{2}\sigma^2S^2\frac{\partial^2V}{\partial S^2})dt+\sigma S\frac{\partial V}{\partial S}dX dV=(∂t∂V+μS∂S∂V+21σ2S2∂S2∂2V)dt+σS∂S∂VdX
Ito 5
针对多因子模型的Ito。有相关性的随机游走。
比如:两只股票的价格 S 1 , S 2 S_1,S_2 S1,S2,价格的SDE形式为: d S i = μ i S i d t + σ i S i d X i dS_i=\mu_iS_idt+\sigma_iS_idX_i dSi=μiSidt+σiSidXi,可知:
d S i 2 = σ 2 S i 2 d t dS^2_i=\sigma^2S^2_idt dSi2=σ2Si2dt
d S 1 d S 2 = ρ σ 1 σ 2 S 1 S 2 d t dS_1dS_2=\rho\sigma_1\sigma_2S_1S_2dt dS1dS2=ρσ1σ2S1S2dt
E [ d X 1 d X 2 ] = ρ d t E[dX_1dX_2]=\rho dt E[dX1dX2]=ρdt
关于S和t的函数形式为: V = V ( t , S 1 , S 2 ) V=V(t,S_1,S_2) V=V(t,S1,S2),时间t是显示存在的。做三维泰勒展开可得:
V ( t + d t , S 1 + d S 1 , S 2 + d S 2 ) = V ( t , S 1 , S 2 ) + ∂ V ∂ t d t + ∂ V ∂ S 1 d S 1 + ∂ V ∂ S 2 d S 2 + 1 2 ∂ 2 V ∂ S 1 2 d S 1 2 + 1 2 ∂ 2 V ∂ S 2 2 d S 2 2 + ∂ 2 V ∂ S 1 ∂ S 2 d S 1 d S 2 V(t+dt,S_1+dS_1,S_2+dS_2)=V(t,S_1,S_2)+\frac{\partial V}{\partial t}dt+\frac{\partial V}{\partial S_1}dS_1+\frac{\partial V}{\partial S_2}dS_2+\frac{1}{2}\frac{\partial^2V}{\partial S^2_1}dS^2_1+\frac{1}{2}\frac{\partial^2V}{\partial S^2_2}dS^2_2+\frac{\partial^2V}{\partial S_1\partial S_2}dS_1dS_2 V(t+dt,S1+dS1,S2+dS2)=V(t,S1,S2)+∂t∂Vdt+∂S1∂VdS1+∂S2∂VdS2+21∂S12∂2VdS12+21∂S22∂2VdS22+∂S1∂S2∂2VdS1dS2
将前面的 d S i , d S i 2 , d S 1 d S 2 dS_i,dS^2_i,dS_1dS_2 dSi,dSi2,dS1dS2代入上式,移项后可得:
d V = ( ∂ V ∂ t + μ 1 S 1 ∂ V ∂ S 1 + μ 2 S 2 ∂ V ∂ S 2 + 1 2 σ 1 2 S 1 2 ∂ 2 V ∂ S 1 2 + 1 2 σ 2 2 S 2 2 ∂ 2 V ∂ S 2 2 + ρ σ 1 σ 2 S 1 S 2 ∂ 2 V ∂ S 1 ∂ S 2 ) d t + σ 1 S 1 ∂ V ∂ S 1 d X 1 + σ 2 S 2 ∂ V ∂ S 2 d X 2 dV=(\frac{\partial V}{\partial t}+\mu_1S_1\frac{\partial V}{\partial S_1}+\mu_2S_2\frac{\partial V}{\partial S_2}+\frac{1}{2}\sigma^2_1S^2_1\frac{\partial^2V}{\partial S^2_1}+\frac{1}{2}\sigma^2_2S^2_2\frac{\partial^2V}{\partial S^2_2}+\rho\sigma_1\sigma_2S_1S_2\frac{\partial^2V}{\partial S_1\partial S_2})dt+\sigma_1S_1\frac{\partial V}{\partial S_1}dX_1+\sigma_2S_2\frac{\partial V}{\partial S_2}dX_2 dV=(∂t∂V+μ1S1∂S1∂V+μ2S2∂S2∂V+21σ12S12∂S12∂2V+21σ22S22∂S22∂2V+ρσ1σ2S1S2∂S1∂S2∂2V)dt+σ1S1∂S1∂VdX1+σ2S2∂S2∂VdX2