Can someone give me an example problem where this particular approach to AI and reasoning hits the ball out of the park? In my mind, it’s difficult to justify a big investment in learning a new subfield without a clear use case where the approach is dramatically superior to other methods.
To be clear, I’m not looking for an example of where the Bayesian approach in general works, I’m looking for an example that justifies the particular strategy of scaling up Bayesian computation, past the point where most analysts would give up, by using MCMC-style inference.
(As an example, deep learning advocates can point to the success of DL on the ImageNet challenge to motivate interest in their approach).
There’s not that many that I know of. I do think its much more intuitive and lets you build more nuanced models that are useful for social sciences. You can fit the exact model that you want instead of needing to fit your case in a preexisting box. However, I don’t know of too many examples where this is hugely practically important.
The lack of obviously valuable use cases is part of why I stopped being that interested in MCMC, even though I invested a lot in it.
There is one important industrial application of MCMC: hyperparameter sampling in Bayesian optimization (Gaussian Processes + priors for hyper parameters). And the hyperparameter sampling does substantially improve things.
Can someone give me an example problem where this particular approach to AI and reasoning hits the ball out of the park? In my mind, it’s difficult to justify a big investment in learning a new subfield without a clear use case where the approach is dramatically superior to other methods.
To be clear, I’m not looking for an example of where the Bayesian approach in general works, I’m looking for an example that justifies the particular strategy of scaling up Bayesian computation, past the point where most analysts would give up, by using MCMC-style inference.
(As an example, deep learning advocates can point to the success of DL on the ImageNet challenge to motivate interest in their approach).
There’s not that many that I know of. I do think its much more intuitive and lets you build more nuanced models that are useful for social sciences. You can fit the exact model that you want instead of needing to fit your case in a preexisting box. However, I don’t know of too many examples where this is hugely practically important.
The lack of obviously valuable use cases is part of why I stopped being that interested in MCMC, even though I invested a lot in it.
There is one important industrial application of MCMC: hyperparameter sampling in Bayesian optimization (Gaussian Processes + priors for hyper parameters). And the hyperparameter sampling does substantially improve things.