move GSOC stuff into subdirectory.

move all html images out of figures folder.

Author: murphyk

figures/amazon-book1-2022-03-01.png amazon-book1-2022-03-01.png

@@ -4,7 +4,7 @@ <h1>Probabilistic Machine Learning: An Introduction</h1>
 	MIT Press, March  2022.
 
 <p>
-<img src="figures/cover1-v2.png"
+<img src="cover1.jpg"
      alt="Book cover"
      style="height:200;">
 <p>
@@ -34,7 +34,7 @@ <h2>Key links</h2>
     <li> <a href="https://github.com/probml/pml-book/tree/main/pml1/README.md">Code to reproduce all the figures</a> 
 
 
-<li> <a href="https://github.com/probml/pml-book/blob/main/figures/transition-guide-2012-to-2022.pdf">Diff from 2012 book</a>
+<li> <a href="https://github.com/probml/pml-book/blob/main/transition-guide-2012-to-2022.pdf">Diff from 2012 book</a>
 
 
 <li> <a href="solns-public.pdf">Solutions to (non-starred) exercises

book1.html

@@ -6,8 +6,8 @@ <h1>Google Summer of Code 2021 - Project "PyProbML"</h1>
 
 
   <p>
-  <img src="figures/cover1.jpg" style="height:200">
-  <img src="figures/cover2.jpg" style="height:200">
+  <img src="../cover1.jpg" style="height:200">
+  <img src="../cover2.jpg" style="height:200">
 
   <p>
 
@@ -108,7 +108,7 @@ <h2><a id="hmm">HMMs for changepoint detection</h2>
 
 
 <p>
-      <img src="figures/hmm-changepoints.png" style="height:300">
+      <img src="hmm-changepoints.png" style="height:300">
 
 
 
@@ -122,9 +122,9 @@ <h2> ><a id="mixfa">Mixtures of (low-rank) Gaussians for CelebA faces</h2>
   After fitting, we can easily generate samples, and perform imputation
   from occluded images,   as shown below.
     <p>
-  <img src="figures/mfa-samples.jpeg" style="height:200">
-  <img src="figures/mfa-occluded.jpeg" style="height:200">
-  <img src="figures/mfa-recon.jpeg" style="height:200">
+  <img src="mfa-samples.jpeg" style="height:200">
+  <img src="mfa-occluded.jpeg" style="height:200">
+  <img src="mfa-recon.jpeg" style="height:200">
 
   <p>
 
@@ -153,7 +153,7 @@ <h2> <a id="potts">Fast blocked Gibbs sampling for the Potts model using JAX</h2
 	<br>
 	This demo took 18 minutes with numpy+numba, but only 6s with JAX+math tricks :).
 <p>
-      <img src="figures/potts-slide.png" style="height:400">
+      <img src="potts-slide.png" style="height:400">
 
 
 
@@ -181,5 +181,5 @@ <h2><a id="cliplogreg">Imagenette classifier using logistic regression + CLIP pr
 
 
 <p>
-      <img src="figures/imagenette-logreg-clip.png" style="height:300">
+      <img src="imagenette-logreg-clip.png" style="height:300">