How does MOE work?¶

Steps toward optimization:

Build a Gaussian Process (GP) with the historical data

Optimize the hyperparameters of the Gaussian Process

Find the point(s) of highest Expected Improvement (EI)

Return the point(s) to sample, then repeat

This section has been given as a 15 minute talk with slides as well.

This section will use examples generated from the Demo Tutorial included with MOE.

Build a Gaussian Process (GP) with the historical data¶

The GP evolves as it is given more historical data, making a better and better model of the underlying function.

See:

Sampling Points

gpp_math

moe.views.rest.gp_mean_var

RW Chapter 2

Optimize the hyperparameters of the Gaussian Process¶

This plot has the same historical data as the plot above. By optimizing the hyperparameters of the GP we select the best model of our underlying function (with respect to a choice of likelihood functions).

See:

Changing Hyperparameters

gpp_covariance

gpp_model_selection

moe.views.rest.gp_hyper_opt

RW Chapter 4

RW Chapter 5

Find the point(s) of highest Expected Improvement (EI)¶

MOE suggests the points of highest Expected Improvement be sampled next. These are the points that are expected to beat the current best sampled point by the most.

See:

gpp_expected_improvement_demo

moe.views.rest.gp_ei

EGO Paper

Return the point(s) to sample, then repeat¶

By continuing to optimize over many iterations, MOE quickly finds approximate optima, or points with large CTR. As the world changes over time, MOE can surf these shifting optima as they move, staying at the peak of the potentially changing objective function in parameter space as time advances.

For more examples on how MOE can be used see Examples, or see Why Do We Need MOE? for information about how MOE is useful for these kinds of problems and some alternate methods.