<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html4/loose.dtd"> <html> <head> <link href="https://bactra.org/plain-style.css" rel="stylesheet" type="text/css"> <title>Generative Diffusion Models</title>        </head> <body>   <P> <P></P><!-- Access to MathJax through the version I'm hosting on bactra.org !--><!-- but only IF mathmode is activated via meta_mathmode !-->  <script> window.MathJax = {  tex: {    inlineMath: [['$','$'], ['\\(','\\)']],    processEscapes: true,    autoload: {      color: [],      colorv2: ['color']    },    packages: {'[+]': ['noerrors']}  },  options: {    ignoreHtmlClass: 'tex2jax_ignore',    processHtmlClass: 'tex2jax_process'  },  loader: {    load: ['[tex]/noerrors']  }};</script>      <cite><a href="http://bactra.org/notebooks">Notebooks</a></cite> <div class="text"><div class="left"><h2>Generative Diffusion Models</h2> <i><a href="faq.html#5.4">Last update</a>: 26 Mar 2026 12:55</i><br><i><a href="faq.html#5.4">First version</a>: 14 April 2025</i> <hr> <P>  <P>Yet Another Inadequate Placeholder <P>for <a href="nn-attention-and-transformers.html">yet another topic I need to master before I become obsolete</a>, this one living where <a href="stoch-diff-eqs.html">stochastic differential equations</a> meets<a href="density-estimation.html">density estimation</a> and <a href="neural-nets.html">neuralnetworks</a>. <ul>Recommended:	<li>Xianghao Kong, Rob Brekelmans, and Greg Ver Steeg, "Information-Theoretic Diffusion", <a href="https://openreview.net/forum?id=UvmDCdSPDOW">ICLR 2023</a>	</ul>  <ul>To read:	<li>Giulio Biroli, Tony Bonnaire, Valentin de Bortoli, Marc M&eacute;zard, "Dynamical Regimes of Diffusion Models", <a href="http://arxiv.org/abs/2402.18491">arxiv:2402.18491</a>	<li>Z. I. Botev, J. F. Grotowski, D. P. Kroese, "Kernel density estimation via diffusion", <a href="https://doi.org/10.1214/10-AOS799"><cite>Annals of Statistics</cite> <strong>38/strong> (2010): 2916--2957</a>, <a href="http://arxiv.org/abs/1011.2602">arxiv:1011.2602</a>	<li>"Count Bayesie", <a href="https://www.countbayesie.com/blog/2023/4/21/linear-diffusion">Linear Diffusion: Building a Diffusion Model from linear Components</a> [I think this would be a nice class exercise, but I need to re-read it]	<li>Reza Ghane, Anthony Bao, Danil Akhtiamov, Babak Hassibi, "Concentration of Measure for Distributions Generated via Diffusion Models", <a href="http://arxiv.org/abs/2501.07741">arxiv:2501.07741</a>	<li>Negar Kamali, Karyn Nakamura, Angelos Chatzimparmpas, Jessica Hullman, Matthew Groh, "How to Distinguish AI-Generated Images from Authentic Photographs", <a href="http://arxiv.org/abs/2406.08651">arxiv:2406.08651</a>	<li>Maxim Raginsky		<ul>		<li><cite>Stochastic Differential Equations: A Systems-Theoretic Approach</cite> [<a href="https://maxim.ece.illinois.edu/teaching/spring23/SDE_book.pdf">PDF draft via Prof. Raginsky</a>.  In this context, for the last chapter.]		<li>"A variational approach to sampling in diffusion processes", <a href="https://cot.mathres.org/archives/2289"><cite>Communications in Optimization Theory</cite <strong>2026:30</strong> (2026): 1--20</a>, <a href="http://arxiv.org/abs/2405.00126">arxiv:2405.00126</a>		</ul>	<li>Pau Rodriguez, Arno Blaas, Michal Klein, Luca Zappella, Nicholas Apostoloff, Marco Cuturi, Xavier Suau, "Controlling Language and Diffusion Models by Transporting Activations", <a href="http://arxiv.org/abs/2410.23054">arxiv:2410.23054</a>	<li>Jascha Sohl-Dickstein, Eric A. Weiss, Niru Maheswaranathan, Surya Ganguli, "Deep Unsupervised Learning using Nonequilibrium Thermodynamics", <a href="http://arxiv.org/abs/1503.03585">arxiv:1503.03585</a>	</ul>  </p> </div></div> <hr> <div class="footer"><P><a href="http://bactra.org/notebooks/generative-diffusion-models.html">permanent link for thisnote</a> &nbsp; <a href="http://bactra.org/notebooks/generative-diffusion-models.rss">RSS feed for thisnote</a></P></div><div class="foot"> <center> <cite><a href="http://bactra.org/notebooks">Notebooks</a></cite>: &nbsp; <a href="http://www.blosxom.com/">Powered by Blosxom</a> </center> </div> </body></html>