Adjoint Sensitive Method

Apply adjoint sensitive method to loss:

From loss=(ztāˆ’ODE(z0))2loss = (z_t - ODE(z_0))^2 to āˆ‚atāˆ‚t=āˆ’atTāˆ‚f(zt,t,Īø)āˆ‚z\frac{\partial a_t}{\partial t} = -a_t^T \frac{\partial f(z_t, t, \theta)}{\partial z} where at=āˆ‚lossāˆ‚zta_t = \frac{\partial loss}{\partial z_t}

When there is a unknown vector u and two known matrix A and B, and we want to compute a product

uTBĀ suchĀ thatĀ AB=Cu^T B \text{ such that } AB = C

If there is a vector v and compute

vTCĀ suchĀ thatĀ ATv=uv^T C \text{ such that } A^Tv = u

Thus

vTC=vTAB=(ATv)TB=uTBv^T C = v^T AB = (A^T v)^T B = u^TB

Turn unknown B and unknown u to one known C and one unknown v

Last updated

Was this helpful?