Metropolis-Hastings In the Real World
Who Needs Machine Learning?
The best way to get a feel for what this model looks like is to pick some parameters and see what the model's predictions are. Try to find a setting for all of these dials such that the probability of seeing the specific points we did see is high.
While you're doing that, think about the struggles a computer might have in estimating these parameters. For example, note how properly setting the variance only makes the model better when the means are already in a good place.
Who Needs Machine Learning?
The best way to get a feel for what this model looks like is to pick some parameters and see what the model's predictions are. Try to find a setting for all of these dials such that the probability of seeing the specific points we did see is high.
While you're doing that, think about the struggles a computer might have in estimating these parameters. For example, note how properly setting the variance only makes the model better when the means are already in a good place.