<?xml version="1.0" encoding="utf-8"?><feed xmlns="http://www.w3.org/2005/Atom" xml:lang="en"><title>SuperconducTED updates</title><subtitle>Progress on the SuperconducTED noise engine, from the students who build it.</subtitle><link href="https://superconducted.github.io/en/updates/atom.xml" rel="self" type="application/atom+xml"/><link href="https://superconducted.github.io/en/updates/" rel="alternate" type="text/html"/><id>https://superconducted.github.io/en/updates/</id><updated>2026-10-17T00:00:00Z</updated><author><name>SuperconducTED contributors</name><uri>https://github.com/SuperconducTED</uri></author><rights>CC BY 4.0</rights><icon>https://superconducted.github.io/favicon-48.png</icon><entry><title>The first ANFIS training step is merged</title><link href="https://superconducted.github.io/en/updates/2026-10-17-first-training-step/" rel="alternate" type="text/html"/><id>https://superconducted.github.io/en/updates/2026-10-17-first-training-step/</id><published>2026-10-17T00:00:00Z</published><updated>2026-10-17T00:00:00Z</updated><summary>The engine can now fit the outputs of its fuzzy rules by ridge least squares, the first half of ANFIS training. Also this week: the project's identity and this website went live.</summary><content type="html">&lt;h2 id="what-merged"&gt;What merged&lt;/h2&gt;&#10;&lt;p&gt;&lt;a href="https://github.com/SuperconducTED/superconducted-noise-engine/pull/114" class="proof"&gt;PR #114&lt;/a&gt; merged on&#10;15 October 2026. It adds the first half of ANFIS training: fitting each rule&amp;rsquo;s linear output by&#10;ridge least squares.&lt;/p&gt;&#10;&lt;ul&gt;&#10;&lt;li&gt;&lt;strong&gt;Both rule types.&lt;/strong&gt; The fit works on the type-1 path and on the interval type-2 path&#10;(Nie-Tan).&lt;/li&gt;&#10;&lt;li&gt;&lt;strong&gt;Checked against planted answers.&lt;/strong&gt; Tests build a 3 × 3 × 3 rule base with known outputs,&#10;and the fit recovers them on both paths.&lt;/li&gt;&#10;&lt;li&gt;&lt;strong&gt;Your units in, your units out.&lt;/strong&gt; Inputs are standardised for the fit and mapped back&#10;exactly to raw units.&lt;/li&gt;&#10;&lt;/ul&gt;&#10;&lt;h2 id="what-it-does-not-do-yet"&gt;What it does not do yet&lt;/h2&gt;&#10;&lt;ul&gt;&#10;&lt;li&gt;&lt;strong&gt;No fuzzy-set training.&lt;/strong&gt; The second half of ANFIS, which moves each set&amp;rsquo;s centre and width,&#10;is the next piece of work.&lt;/li&gt;&#10;&lt;li&gt;&lt;strong&gt;No run on the archive.&lt;/strong&gt; Training needs 1,170 &lt;a href="https://github.com/SuperconducTED/superconducted-noise-engine/blob/main/docs/numerical-claims.md" class="proof"&gt;NC-045&lt;/a&gt; distinct device states;&#10;868 &lt;a href="https://github.com/SuperconducTED/superconducted-noise-engine/blob/main/docs/numerical-claims.md" class="proof"&gt;NC-069&lt;/a&gt; are archived so far.&lt;/li&gt;&#10;&lt;/ul&gt;&#10;&lt;h2 id="also-this-week"&gt;Also this week&lt;/h2&gt;&#10;&lt;ul&gt;&#10;&lt;li&gt;&lt;strong&gt;An identity.&lt;/strong&gt; The organisation adopted its mark and colours on 16 October 2026&#10;(&lt;a href="https://github.com/SuperconducTED/.github/pull/1" class="proof"&gt;PR #1&lt;/a&gt; in the organisation repository).&lt;/li&gt;&#10;&lt;li&gt;&lt;strong&gt;This website.&lt;/strong&gt; It went live the same day&#10;(&lt;a href="https://github.com/SuperconducTED/superconducted.github.io/pull/1" class="proof"&gt;PR #1&lt;/a&gt; in the site&#10;repository). Every number on it links to the register row it comes from.&lt;/li&gt;&#10;&lt;/ul&gt;&#10;</content></entry></feed>