JNJosé Nunesinml-classroom.hashnode.dev·Sep 1 · 22 min readProduct of ExpertsIn this article, you will learn what a Product of Experts (PoE) is, why it matters, and how it works. By the end, you will understand the relevance of PoE and be able to understand implementations. P20
CCalculatorQueenincalculatorqueen.hashnode.dev·Aug 28 · 7 min readExpected Value Is Not a Forecast: Engineering a Transparent Crafting CalculatorRandomized crafting systems invite a deceptively simple question: “Is this recipe worth attempting?” A calculator can produce a number quickly, but that number is only useful if its assumptions remain00
UOUche Ozoemenainincodethismeans.com·Jul 16 · 9 min readThe Sigmoid Function: A Layman's IntuitionThe question What does it mean when someone says the sigmoid function converts outputs of a predictive model into a probability value? The task To address that question, let's start with a task that's00
FDFundacja Dobre Państwoinfundacjadobrepasntwo.hashnode.dev·May 26 · 1 min readHow Much Does Tomorrow Cost? The Mathematics and Ethics of Financial MarketsThis article provides a deep analysis of the mathematical framework of contemporary finance, treating quantitative tools as a grammar for describing an uncertain future. The author focuses on the tria00
MMacaulay001inforgottentheorieshashnodedev.hashnode.dev·May 11 · 7 min readBerkeley's 1973 admit rates flip sign when you stop pooling the departmentsPart 74 of Forgotten Theories, a series re-testing old scientific claims with modern tools. Find the rest under the #forgotten-theories tag. In 1903 Udny Yule wrote a short paper on the association between two attributes in a 2x2 table. He noticed t...00
MMacaulay001inforgottentheorieshashnodedev.hashnode.dev·May 11 · 6 min readI tested Bertrand's 1887 ballot formula on 648 vote splits and the answer was always (p-q)/(p+q).Part 100 of Forgotten Theories, a series re-testing old scientific claims with modern tools. Find the rest under the #forgotten-theories tag. In 1887 Joseph Bertrand sent a four-page note to the Comptes Rendus of the French Academy. The question he ...00
MMacaulay001inforgottentheorieshashnodedev.hashnode.dev·May 11 · 6 min readI tested von Mises's 1939 birthday claim with exact arithmetic and 10^5 Monte Carlo trials per group size. 23 people really is enough.Part 99 of Forgotten Theories, a series re-testing old scientific claims with modern tools. Find the rest under the #forgotten-theories tag. In 1939 Richard von Mises published a short paper with the unwieldy title Über Aufteilungs- und Besetzungswa...00
MMacaulay001inforgottentheorieshashnodedev.hashnode.dev·May 11 · 6 min readChebyshev's 1867 bound holds on every distribution I threw at it, and a two-point mix landed within 0.2% of saturating it.Part 98 of Forgotten Theories, a series re-testing old scientific claims with modern tools. Find the rest under the #forgotten-theories tag. In 1867 Pafnuty Chebyshev published a short proof of an inequality that any student of probability eventuall...00
MMacaulay001inforgottentheorieshashnodedev.hashnode.dev·May 11 · 6 min readI sampled a million coin flips with decaying probabilities. The 1909 dividing line showed up exactly where Borel said it would.Part 96 of Forgotten Theories, a series re-testing old scientific claims with modern tools. Find the rest under the #forgotten-theories tag. In 1909 Émile Borel was working through a problem about decimal expansions of real numbers, and he proved a ...00
MMacaulay001inforgottentheorieshashnodedev.hashnode.dev·May 11 · 7 min readI ran Paul Levy's 1925 generalized CLT on 540,000 sums of heavy-tailed variables and the stable laws matched to KS = 0.005.Part 79 of Forgotten Theories, a series re-testing old scientific claims with modern tools. Find the rest under the #forgotten-theories tag. The central limit theorem is the most quoted result in statistics. Add up many independent things, divide by...00
