previous up next print clean
Next: About this document ... Up: Claerbout: CG Huber regression Previous: PREDICTION-ERROR FILTER ESTIMATION

CONCLUSION

I feel the method of this paper did not work very well. I believe I should prepare and test a code for the Huber line search and then use standard conjugate gradients to see if that works better than the conjugate-direction method.



Stanford Exploration Project
11/12/1997