Nullspace shuttling on an Ising network

Kristopher A. Innanen

In a companion report, AVO inversion for general parameterization is formulated and examined numerically in the context of Ising computation. Here the capacity of the Ising model to further help explore the so-called inversion nullspace as a means to quantify un- certainty is developed. This involves Monte Carlo sampling along similar lines as the inverse workflow, with two differences: (1) that we consider bit exchanges instead of flips of bit states, and (2) we accept candidate exchanges preferentially if the energy is main- tained, and penalize increases and decreases. Numerical examination reveals that the Ising nullspace exploration algorithm works, and does recover the expected tradeoff between P - wave velocity and density perturbations, but fails to identify that the Ising inversion has poorly estimated the gradient term.