Iteration Dogma: Difference between revisions
From Rest of What I Know
Created page with "The Iteration Dogma is the belief that adapting to evidence is far superior to picking the right prior. It is commonly employed in the startup-mindset of searching for Product-market fit where it is considered boringly true. == Examples == * Searching for product-market fit is the primary example. * Bayesian updating under prior-support, distinguishability of hypotheses (particularly truth from alternatives in the wikipedia:Kullb..." |
Add popper |
||
| Line 6: | Line 6: | ||
* Bayesian updating under prior-support, distinguishability of hypotheses (particularly truth from alternatives in the [[wikipedia:Kullback–Leibler divergence|KL divergence]] sense) | * Bayesian updating under prior-support, distinguishability of hypotheses (particularly truth from alternatives in the [[wikipedia:Kullback–Leibler divergence|KL divergence]] sense) | ||
* [[wikipedia:Karl Popper|Popper]]-style empirical science with [[wikipedia:Critical rationalism|falsifiable conjectures formed which are iteratively refined through targeted refutation]] | |||
== Where It Does Not Work == | == Where It Does Not Work == | ||
Revision as of 23:29, 27 January 2026
The Iteration Dogma is the belief that adapting to evidence is far superior to picking the right prior. It is commonly employed in the startup-mindset of searching for Product-market fit where it is considered boringly true.
Examples
- Searching for product-market fit is the primary example.
- Bayesian updating under prior-support, distinguishability of hypotheses (particularly truth from alternatives in the KL divergence sense)
- Popper-style empirical science with falsifiable conjectures formed which are iteratively refined through targeted refutation
Where It Does Not Work
The iteration dogma is a bad idea to employ in fields where global convergence are not possible. e.g. finding the roots of a polynomial employing Newton-Raphson will result in slow convergence if in the wrong place.
