UQ_SRMV_Marmousi

1. Run RSVD_SRVM.m to get the SVD of inverse Hessian

2. Run sampling_Hess to extract elements from Hessian or inverse Hessian

3. Run T_distribute.m to get the square-root of inverse Hessian, in which 1000 2d random samplers are generated.

4. DisTopV.m: posterior distribution

5. DisTopV0.m: prior distribution

6. Hist_curve.m: visualized depth Histogram profiles

7. plotsub.m, plotsub0.m: visualized marginal distributions

8. null_spac.m: null space analysis

Others: supporting subfunctions

The SRVM scalars a_* and nu_, and vectors w_ can be collected from the ./scratch/optimize folder under the numerical example.

Actually, only one scalar variable $ nu/a $ suffices, which will occur in the official version soon.

The initModel folder contains the init models, used for prior samplings.

The invertedModel folder contains the inverted models, used for posterior sampling.

Note that the Misfit functions in SRVM_part1 were obtained 6 months ago.

For the newest comparion of data misfit functions, please find data_misfit.png within this folder.

Please take data_misfit_old.png for reference.