2019-06-24 - Bayesian vs Frequentist

 

~5.3 mi @ ~12.8 min/mi

"No Free Lunch theorem" - "Pigeonhole principle" - "Shattered sets" - "Curry's paradoxical combinator" - The jargon is strong this morning, as Keymaster attempts to explain theoretical foundations of machine learning to Roadkill, who tries to map mind-altering new concepts to old. K2 simply smiles, an appropriate Zen commentary on the entire enterprise!

"That looks just like my grandmother's home - but many times larger!" Dawn Patrol explores the lovely-quiet neighborhood cuddled between Scotts Run Nature Preserve, the Capital Beltway, and the Potomac River, where sprawling mansions take dominion over slovenly wilderness. A new eagle statue spreads its wings above a courtyard.

"Eat our Bubbles!" reads graffiti on the local swim club driveway, beginning trash talk to visiting teams before they even get out of their cars. A rabbit freezes, then dashes away; a deer strides through the brush.

"Bad luck is followed by good!" K2 reports that her kids observe Regression Toward the Mean in their travel misadventures and recoveries therefrom. Roadkill explains why a Bayesian in trouble would pick a random person to beg help of, rather than accept aid from a self-selected Samaritan. Keymaster spies a fox, or thinks he does. When multiple cats are then seen in a nearby yard he concedes that some probabilities need adjustment. "So with what confidence level are you a Frequentist?" Hmmmm ...

(trackfile) - ^z - 2019-07-14