cs401r_w2016:lab4
Differences
This shows you the differences between two versions of the page.
| Both sides previous revision Previous revision Next revision | Previous revision | ||
|
cs401r_w2016:lab4 [2015/12/23 23:44] admin |
cs401r_w2016:lab4 [2021/06/30 23:42] (current) |
||
|---|---|---|---|
| Line 3: | Line 3: | ||
| To understand Gaussian process regression, and to be able to generate nonparametric regressions with confidence intervals. Also to understand the interplay between a kernel (or covariance) function, and the resulting confidence intervals of the regression. | To understand Gaussian process regression, and to be able to generate nonparametric regressions with confidence intervals. Also to understand the interplay between a kernel (or covariance) function, and the resulting confidence intervals of the regression. | ||
| + | ---- | ||
| ====Deliverable:==== | ====Deliverable:==== | ||
| Line 16: | Line 17: | ||
| + | ---- | ||
| + | ====Grading standards:==== | ||
| + | |||
| + | |||
| + | Your notebook will be graded on the following: | ||
| + | |||
| + | * 20% Correct implementation of three kernels | ||
| + | * 30% Correct implementation of noiseless GPR | ||
| + | * 30% Correct implementation of noisy GPR | ||
| + | * 20% Six tidy and legible plots, with appropriate ranges | ||
| + | |||
| + | ---- | ||
| ====Description:==== | ====Description:==== | ||
| Line 25: | Line 38: | ||
| </code> | </code> | ||
| - | You must perform Gaussian process regression on this dataset, and produce visualizations for both noiseless and noise-free observations. Your notebook should produce one visualization for each of the following kernel types: | + | You must perform Gaussian process regression on this dataset, and produce visualizations for both noiseless and noisy observations. Your notebook should produce one visualization for each of the following kernel types: |
| * The linear kernel (MLAPP 14.2.4) | * The linear kernel (MLAPP 14.2.4) | ||
| Line 33: | Line 46: | ||
| Therefore, your notebook should produce **six** different visualizations: two for each kernel type. | Therefore, your notebook should produce **six** different visualizations: two for each kernel type. | ||
| - | For the noisy observation case, use \sigma_n^2=0.1 | + | For the noisy observation case, use \sigma_n^2=0.1. |
| + | |||
| + | For the polynomial kernel use a degree of 3. | ||
| + | |||
| + | For the Gaussian / RBF kernel, set all parameters to 1.0 | ||
| + | |||
| + | The mean function for this lab should always return 0. | ||
| You should also answer the following questions: | You should also answer the following questions: | ||
| Line 40: | Line 59: | ||
| * What happens when the degree M of the polynomial kernel gets small? Gets large? | * What happens when the degree M of the polynomial kernel gets small? Gets large? | ||
| + | Your visualizations should be done on the range ''[-2 10]'' of the x-axis. | ||
| + | |||
| + | For the errorbars, you can just plot the mean +/- the variance. This isn't really a statistically meaningful quantity, but it makes the plots look nice. :) | ||
| + | |||
| + | //Hint: a Gaussian process only allows you to make a prediction for a single query point. So how do you generate the smoothly varying lines in the example images?// | ||
| + | |||
| + | ---- | ||
| ====Hints:==== | ====Hints:==== | ||
| Line 45: | Line 71: | ||
| <code python> | <code python> | ||
| + | |||
| + | numpy.arange() | ||
| plt.gca().fill_between | plt.gca().fill_between | ||
| plt.scatter | plt.scatter | ||
| - | numpy.linalg.inv | + | numpy.linalg.pinv |
| numpy.eye | numpy.eye | ||
cs401r_w2016/lab4.1450914279.txt.gz · Last modified: 2021/06/30 23:40 (external edit)
Page Tools
Except where otherwise noted, content on this wiki is licensed under the following license: CC Attribution-Share Alike 4.0 International
