add notes back

index.html

@@ -84,7 +84,7 @@ <h2>Recent Publications</h2>
 
 <p>
 Finally, I have a collection of brief, tutorial-style <a href="notes/">research
-notes</a>.
+notes</a> (very old).
 </p>
 
 

notes/index.html

@@ -0,0 +1,52 @@
+---
+layout: default
+title: "Tutorial Notes: Charles Sutton"
+---
+
+<h1>Research-ish Notes</h1>
+
+<p>
+Inspired by the example of 
+<a href="http://people.csail.mit.edu/jrennie/writing/">Jason 
+Rennie</a>, I wrote some short notes to help me learn background
+during my PhD.  Generally, these are 2- or 3-page 
+simple notes that answer a question either about my work
+or someone else's.  Others of these describe implementation
+tricks that may be
+useful for efficiently reproducing my or others' results.
+They are pretty old now, but perhaps they are still useful
+to someone.
+</p>
+
+
+<ul>
+  <li>
+      <b><a href="sutton06backprop.pdf">CRF Gradients Without Forward Passes</a></b>.
+      Charles Sutton, April 2006
+  </li>
+  <li>
+  <b><a href="sutton05tree-kl.pdf">Factorization of KL-divergence Between Tree-Structured Distributions</a></b>.
+    Charles Sutton, July 2005
+  </li>
+  <li>
+<b><a href="sutton04fbbp.pdf">Understanding Why Forward-Backward is Belief Propagation in Two Easy Steps</a></b>.
+Charles Sutton, June 2004
+  </li>
+</ul>
+
+<p>
+If you find one of these notes useful, and wish to cite it,
+please use a citation like the following:
+</p>
+<pre>
+@misc{sutton04understanding,
+  author = "Charles Sutton",
+  title = "Understanding Why Forward-Backward is Belief Propagation in Two Easy Steps",
+  year = 2004,
+  month = "June",
+  howpublished = "http://homepages.inf.ed.ac.uk/csutton/notes/"
+}
+</pre>
+
+<p>(The dates on these notes are the dates that
+I last revised them.)</p>