Differences
This shows you the differences between the selected revision and the current version of the page.
| sep:courses:gp211_labs 2009/10/23 13:33 | sep:courses:gp211_labs 2009/11/15 23:27 current | ||
|---|---|---|---|
| Line 18: | Line 18: | ||
| | {{sep:courses:gif:lab4.gif}} | **Missing Data Estimation\\ and Multiple Realizations **\\ \\ {{sep:courses:labs211:labs:lab4:lab4.pdf|Lab4.pdf}}| Geostatistics has gained much recent attention for its ability to solve missing data problems not with a single answer, but with a suite of so-called equiprobable models, or models that fit the data equally well. The geostatistical approach gives better results in fluid flow simulation (for instance), and also provides a useful tool for multiple scenario analysis. Clapp (SEP-105) showed that multiple realizations can easily be incorporated into the existing SEP framework to solving missing data problems. This lab has two parts. In the first, you will modify a conjugate gradient algorithm to allow multiple realizations, and then test it on the 1-D inverse linear interpolation problem given in Chapter 3 of the book. In the second part, you will use your new solver to tackle the ** Wells not matching the seismic map ** problem of Chapter 3. You will improve the ``extension'' of the seismic data into empty regions of the model space, first by writing a program to de-trend the seismic data with a least squares best-fitting plane, and second, by using your new solver to come up with multiple realizations of the extended seismic data. | {{sep:courses:labs211:labs:lab4:lab4.tar|Lab4.tar}}| | | {{sep:courses:gif:lab4.gif}} | **Missing Data Estimation\\ and Multiple Realizations **\\ \\ {{sep:courses:labs211:labs:lab4:lab4.pdf|Lab4.pdf}}| Geostatistics has gained much recent attention for its ability to solve missing data problems not with a single answer, but with a suite of so-called equiprobable models, or models that fit the data equally well. The geostatistical approach gives better results in fluid flow simulation (for instance), and also provides a useful tool for multiple scenario analysis. Clapp (SEP-105) showed that multiple realizations can easily be incorporated into the existing SEP framework to solving missing data problems. This lab has two parts. In the first, you will modify a conjugate gradient algorithm to allow multiple realizations, and then test it on the 1-D inverse linear interpolation problem given in Chapter 3 of the book. In the second part, you will use your new solver to tackle the ** Wells not matching the seismic map ** problem of Chapter 3. You will improve the ``extension'' of the seismic data into empty regions of the model space, first by writing a program to de-trend the seismic data with a least squares best-fitting plane, and second, by using your new solver to come up with multiple realizations of the extended seismic data. | {{sep:courses:labs211:labs:lab4:lab4.tar|Lab4.tar}}| | ||
| - | | {{sep:courses:gif:lab5.gif}} | ** Interval\\ Velocity\\ Estimation **\\ \\ | This lab will guide you through most of the steps - and pitfalls - of interval velocity estimation, perhaps the most important problem in exploration geophysics. You will apply the least squares velocity scan of Lab 2, and then pick RMS velocities using an autopicker and by hand. You will apply the multiple realizations concept to bound uncertainty on your picked RMS velocity. You will solve the Dix equation as a least squares problem to estimate interval velocity in 1-D and make some preliminary steps in the direction of 2-D. | | + | | {{sep:courses:gif:lab5.gif}} | ** Interval\\ Velocity\\ Estimation **\\ \\ {{sep:courses:labs211:labs:lab5:lab5.pdf|Lab5.pdf}}| This lab will guide you through most of the steps - and pitfalls - of interval velocity estimation, perhaps the most important problem in exploration geophysics. You will apply the least squares velocity scan of Lab 2, and then pick RMS velocities using an autopicker and by hand. You will apply the multiple realizations concept to bound uncertainty on your picked RMS velocity. You will solve the Dix equation as a least squares problem to estimate interval velocity in 1-D and make some preliminary steps in the direction of 2-D. | {{sep:courses:labs211:labs:lab5:lab5.tar|Lab5.tar}} | |
| - | | {{sep:courses:gif:lab6.gif}} | **Debluring \\ by inversion \\ with regularization**\\ \\ | In this Lab, we test different regularization schemes for deblurring a text image with random noise. Regularization is very important in inversion when there are not enough data. Prior information can be included in the regularization to make the inversion converge fast. In this Lab, you will code the adjoint of the given operator. You will then apply two regularization operators to the linear inversion. In the final part, you will apply an edge-preserving regularization scheme to the non-linear inversion. | | + | | {{sep:courses:gif:lab6.gif}} | **Debluring \\ by inversion \\ with regularization**\\ \\ {{sep:courses:labs211:labs:lab6:lab6.pdf|Lab6.pdf}} | In this Lab, we test different regularization schemes for deblurring a text image with random noise. Regularization is very important in inversion when there are not enough data. Prior information can be included in the regularization to make the inversion converge fast. In this Lab, you will code the adjoint of the given operator. You will then apply two regularization operators to the linear inversion. In the final part, you will apply an edge-preserving regularization scheme to the non-linear inversion. | {{sep:courses:labs211:labs:lab6:lab6.tar|Lab6.tar}} | |
| - | | {{sep:courses:gif:lab7.gif}} | ** Madagascar **\\ \\ | The Madagascar satellite data set provides images of a spreading ridge off the coast of Madagascar. This data set has two regions: the southern half is densely sampled and the northern half is sparsely sampled. The sparsely sampled region presents a missing data problem. In this exercise, you will apply basic fitting goals to image the dense area. Then, you will be asked to fill the missing data in th --- //[[sjoerd@sep.stanford.edu|Sjoerd de Ridder]] 2009/10/15 23:55//e sparsely sampled region. Additionally, you will use prediction-error filters estimated on the dense southern half to fill data on the sparse half. Preconditioning will be used to speed convergence. The prediction-error filters will attempt to spread the texture of the spreading ridge to the sparse tracks. | | + | | {{sep:courses:gif:lab7.gif}} | ** Madagascar **\\ \\ | The Madagascar satellite data set provides images of a spreading ridge off the coast of Madagascar. This data set has two regions: the southern half is densely sampled and the northern half is sparsely sampled. The sparsely sampled region presents a missing data problem. In this exercise, you will apply basic fitting goals to image the dense area. Then, you will be asked to fill the missing data in the sparsely sampled region. Additionally, you will use prediction-error filters estimated on the dense southern half to fill data on the sparse half. Preconditioning will be used to speed convergence. The prediction-error filters will attempt to spread the texture of the spreading ridge to the sparse tracks. | |
| - | | LOL | ** Extra **\\| {{sep:courses:labs211:labs:extra:extra.pdf|Extra.pdf}}| An additional assignment designed by Jeff Shragge, to aid you in obtaining coding skills. Have fun! |{{sep:courses:labs211:labs:extra:extra.tar|Extra.tar}}| | + | | LOL | ** Extra ** | {{sep:courses:labs211:labs:extra:extra.pdf|Extra.pdf}}| An additional assignment designed by Jeff Shragge, to aid you in obtaining coding skills. Have fun! |{{sep:courses:labs211:labs:extra:extra.tar|Extra.tar}}| |
| \\ \\ | \\ \\ | ||
| {{page>share:footer&nofooter&noeditbtn}} | {{page>share:footer&nofooter&noeditbtn}} | ||
