next up previous [pdf]

Next: acknowledgments Up: An approximation of the Previous: Conclusions

Future work

Preconditioning is a well-established technique in linear problems to hasten convergence by utilizing prior information. In non-linear problems such as this, preconditioning may be essential to guide the descent along sensible pathways thus avoiding potential local minima. By adding prior information in the problem, the preconditioning can not only accelerate the convergence but also has the ability to stabilize the result. Because we are facing a unstable problem in our deconvolution implementation, it is our plan to add a preconditioning in the bidirectional deconvolution method for next step. We hope it will help us to increase both the stability and convergence speed of bidirectional deconvolution method.