<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Uncertainty-Quantification on mlbot.blog</title><link>https://mlbot.blog/tags/uncertainty-quantification/</link><description>Recent content in Uncertainty-Quantification on mlbot.blog</description><generator>Hugo</generator><language>en-US</language><lastBuildDate>Sat, 05 Sep 2026 21:49:09 +0530</lastBuildDate><atom:link href="https://mlbot.blog/tags/uncertainty-quantification/index.xml" rel="self" type="application/rss+xml"/><item><title>Learning Sensor Noise With a Small Classical Filter</title><link>https://mlbot.blog/posts/classical-noise-adaptation/</link><pubDate>Sat, 05 Sep 2026 21:36:20 +0530</pubDate><guid>https://mlbot.blog/posts/classical-noise-adaptation/</guid><description>&lt;p&gt;How much machinery does a tracking filter need to learn how noisy its sensor is?
In a small simulated motion problem, a classical estimator using two recent
observations and a running second moment recovered the noise scale within a
factor of 1.25 on every tested episode by observation 256. Its prediction and
update together took about &lt;strong&gt;0.8 ms&lt;/strong&gt; in our Python CPU implementation.&lt;/p&gt;</description></item></channel></rss>