"In this age of intelligent machines... Bayesian methods have an important and novel role to play."
"Bayesian statistics adds something a little new."
"Bayes rule gives us a way to move between probability of data given model or in probability of model given data."
"We draw a number from our prior distributions, we assume that to be true and now fit it, put that into the likelihood."
"To be a good Bayesian is really about trying your best to admit that you have priors and to update them when you get new information as best you can."
"Bayesian deep learning is basically grounded on learning probability distributions over our parameters rather than point estimates."
"The nice thing about being Bayesian is that if there's anything you don't know about the model even if it's a nuisance parameter what you do is you put a prior on and add it to the model and estimate it just like everything else."
"With the Bayesian test we are not just dealing with two point estimates. The output of the Bayesian test is two distributions."
"The Bayesian view takes a completely different approach."
"It supports everything that GLM supports but adds some additional arguments due to the fact that we're going to use Bayesian estimation."
"In Bayesian modeling, the full posterior distribution provides a rich understanding of uncertainty, including parameter dependencies and propagation of uncertainty to derived quantities."
"There's not one true line in Bayesian analysis. There's no one point from the posterior distribution or one line that is the right line or the right answer. It's the whole distribution."
"A Bayesian approach always uses probability as the system of reasoning."
"This is a particular characteristic of Bayesian estimation that allows us to make probabilistic statements about unknowns conditional on the data."
"Bayesian inference obeys the likelihood principle that says all information relevant to the unknown parameters in your model are contained in the likelihood."
"In the Bayesian framework, it allows you to incorporate prior knowledge."
"The biggest pro of Bayesian thinking is that it simulates the way human beings think about the world."
"Scientists act more Bayesian, we try to pile up evidence in favor of the theory and the more you have the more likely you are to gain acceptance in your community of scientists."
"Point estimates don't really have a big role in Bayesian inference because the estimate is the curve."
"The big advantage of Bayes is that you can take the posterior from one analysis and then use it as the prior for the next time you do an analysis."
"Bayesian methods for hackers... gives you really intuitive taste of Bayesian inference."
"We're not Bayesian updaters; that is, we're not ideal Bayesian agents."
"Bayesian methods and techniques are incredibly valuable in a whole range of cases."
"The nice thing about Bayesian statistics is it allows you to incorporate prior beliefs into your estimation process."
"You need to your prior probability, then you get the result and you adjust your prior probability afterwards."
"We want to use all possible settings of parameters and weight them by their posterior probabilities in what's called a Bayesian model average."
"With the Bayesian approach, we can include a prior distribution for the model parameters and then calculate the likelihood from the data in order to sample the posterior for the model parameters."
"The Bayesian approach comfortably accommodates the idea that you're going to have to bring in some information from outside the model."
"Instead of saying 'Hey, I'm uncertain about my data but I'm certain about my model,' Bayesians prefer to say 'Well, I'm a little bit uncertain about my model, actually, I'm quite certain about the data.'"
"The Bayesian goes to the frequentist and says, 'Oh, you're very average.' And the Bayesian says, 'No, you're mean.'"
"It would be better to use Bayesian approaches in presenting and interpreting results."
"Bayesian optimization... we use uncertainties to decide the trade-off between exploration versus exploitation."
"If you want to do a robust analysis where you want to put all your prior assumptions and verify that those priors, that intuitions are right by looking at data and predictive tests, then the Bayesian scheme is ideal for that."
"Bayesians in particular are not just friendly but they're very polite."
"The nice thing about the Bayesian framework is it allows us to recover the usual common sense estimators just based on the data."
"The predictive distribution is the distribution that a Bayesian would want."
"Minimizing Bayesian regret can actually be reduced pretty exactly to minimizing the Bayesian anytime motion planning objective."
"For Bayesian learning, there are two things that we need to specify: the data model or what we often call the likelihood, and we need to specify a prior."
"I'm Bayesian and what I want to do is I want to consider the parameters which are most likely given the data."
"Bayesian networks provide a graphical representation."
"Most people desperately want to be Bayesians; they just aren't aware of it."
"Using Bayes opens up other possibilities that we just didn't have when we stuck with maximum likelihood."
"The Bayesian framework... treats the underlying parameter H as random in nature."
"We would like to use the Bayesian or the minimum mean squared error principle for parameter estimation."
"Posterior is equal to likelihood times prior divided by the evidence."
"When you do Bayesian inference, you have to introduce a prior on your parameter."
"The Bayesian School says, 'We don't know what the value of theta is, so let's represent our uncertainty over theta with a prior.'"