<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Gsoc26s on Gökhan Kof</title><link>https://kofgokhan.github.io/gsoc26/</link><description>Recent content in Gsoc26s on Gökhan Kof</description><generator>Hugo</generator><language>en-us</language><lastBuildDate>Fri, 15 May 2026 17:24:49 +0300</lastBuildDate><atom:link href="https://kofgokhan.github.io/gsoc26/index.xml" rel="self" type="application/rss+xml"/><item><title>GSoC '26</title><link>https://kofgokhan.github.io/gsoc26/gsoc/</link><pubDate>Fri, 15 May 2026 17:24:49 +0300</pubDate><guid>https://kofgokhan.github.io/gsoc26/gsoc/</guid><description>&lt;p&gt;I will post any updates here on this page.&lt;/p&gt;
&lt;p&gt;I might add more details later in order to make these more useful than just a simple log of activities.&lt;/p&gt;
&lt;h2 id="week-1-2"&gt;Week 1-2&lt;/h2&gt;
&lt;p&gt;So far I have a good handle on how the package MathOptAI.jl works and how the workflow of implementing a new predictor should be handled. I have implemented &lt;code&gt;ReLUEpigraph&lt;/code&gt; and currently testing it on a practical problem.&lt;/p&gt;
&lt;p&gt;I have also developed a &lt;code&gt;Flux&lt;/code&gt; model which is an Input Convex Neural Network (ICNN) that can be embedded in a &lt;code&gt;JuMP.Model&lt;/code&gt; and using &lt;code&gt;ReLUEpigraph&lt;/code&gt; enables the ICNN to be represented with an LP formulation. The PR for the &lt;code&gt;ReLUEpigraph&lt;/code&gt; is created &lt;a href="https://github.com/lanl-ansi/MathOptAI.jl/pull/274"&gt;here&lt;/a&gt;.&lt;/p&gt;</description></item></channel></rss>