Angie(Hyunji) Moon

@mcmc_stan developer, Bayesian and optimization enthusiast, BDA3 translator, CEO of NextOpt

[Angie(Hyunji) Moon @bayesmoon]

@LeftyVi Thanks for asking! It’ll be shared in https://t.co/JB7tuOQJLH. Feedback welcome on the last year’s slides so that we can update this year!

[Angie(Hyunji) Moon @bayesmoon]

🔬 Researchers at #AOM2024!

Don't miss Bayesian Statistics: Conduct & Publish High-Quality Studies
to learn
• Bayesian vs. significance tests
• Step-by-step studies
• Priors, MCMC, communication

Join Bayesian community! Resources available. Register: https://t.co/lds81R3yot

[Angie(Hyunji) Moon @bayesmoon]

@LeftyVi i question that myself 🤷‍♀️ but FWIW conference spans from Friday to Wednesday

[Peyman Milanfar @docmilanfar]

Why don’t more people know about the gem that is Tweedie's formula?

Say 𝐱 is a noisy measurement of 𝐮

𝐱 = 𝐮 + 𝐞

w/ 𝐞 ∼ 𝒩(𝟎, σ² 𝐈)

min mean² estimate of 𝐮 is 𝔼 [𝐮 | 𝐱]. Obviously we need the density P(𝐮|𝐱) right?

No! Tweedie says P(𝐱) is all you need!

1/2

[Angie(Hyunji) Moon @bayesmoon]

@sp_monte_carlo @betanalpha 🌗🌓 My favorite word ‘Bayequentist’ :) invented from ‘The Possimpible’
(How I Met Your Mother: Season 4, Episode 14)

[Angie(Hyunji) Moon @bayesmoon]

@betanalpha I use cmdstanpy and arviz. Arviz Xarray data type is useful to me.

[Angie(Hyunji) Moon @bayesmoon]

@betanalpha Why So Serious? Machine learning you complete me - Bayesian Joker

[Angie(Hyunji) Moon @bayesmoon]

@Adam235711 Time delay :)

[Angie(Hyunji) Moon @bayesmoon]

Excited to get feedback on the simulation library I am building 🙃

[Angie(Hyunji) Moon @bayesmoon]

@MakkarAditya Quote is from Doob’s textbook Stochastic processes!

[Angie(Hyunji) Moon @bayesmoon]

Prior modeling can be viewed as subjective optimal transport.

[Angie(Hyunji) Moon @bayesmoon]

@betanalpha @GuidoBiele @jbakcoleman I view SBC’s role parallel to that of defensive prior. Lesser evil, opt out. May I go one step further to argue “SBC:Valid workflow ~ KKT:Local optimum”? Hardly an iff relation which doesn’t detract KKT’s status as a cheap optimum litmus (with care)!

[Angie(Hyunji) Moon @bayesmoon]

@betanalpha @GuidoBiele @jbakcoleman I wish to question the need to test individual posterior when our inference would not be based on its individual version, but rather on its pooled one. SBC as a function of experimental settings would help gauge reliability closet to granularity of its actual use.

[Angie(Hyunji) Moon @bayesmoon]

@thatpinkney @searchbooster_ Congratulations!

[Angie(Hyunji) Moon @bayesmoon]

@betanalpha @GuidoBiele @jbakcoleman I view SBC’s role parallel to that of defensive prior. Lesser evil, opt out. May I go one step further to argue “SBC:Valid workflow ~ KKT:Local optimum”? Hardly an iff relation which doesn’t detract KKT’s status as a cheap optimum litmus (with care)!

[Angie(Hyunji) Moon @bayesmoon]

@betanalpha @GuidoBiele @jbakcoleman I wish to question the need to test individual posterior when our inference would not be based on its individual version, but rather on its pooled one. SBC as a function of experimental settings would help gauge reliability closet to granularity of its actual use.

[Angie(Hyunji) Moon @bayesmoon]

I am honored and excited to be TAing for prof. David Yao’s stochastic modeling class! 🏛 https://t.co/DWYrfrjz43

[Angie(Hyunji) Moon @bayesmoon]

@sp_monte_carlo Thanks, I wonder what might be the insight behind “symmetric distribution taking three possible values −√a, 0, √a, s.t. probability of a non-zero value is 1/a”. Three from knowing determining geometry.. Might 1/a * (√a)^2 =1 mean anything?

[Angie(Hyunji) Moon @bayesmoon]

@schmaidt Thanks for the reply. I am sorry but could you please elaborate a little on the first proof?