progress with unsw slides, part 1

talks/2024_UNSW_CSE_Alex.html

@@ -93,12 +93,16 @@ <h3>The epistasis model</h3>
 	$$e = f(00) + f(11) - \Big(f(01) + f(10)\Big)$$
 
 	<aside class="notes">
-	  Mention shapes
+	  Mention shapes, higher-order interaction detection, TCS, etc.
 	</aside>
 </section>
 
 <section>
 	<img data-src="/assets/fitlands_outline.png" width=60%>
+
+	<aside class="notes">
+	  This is SotA pre AI in Bio revolution -- AF(1) came out in 2018.
+	</aside>
 </section>
 
 <section>
@@ -109,6 +113,11 @@ <h4>State of the art in 2018</h4>
 		Kieran Elmes et al. Learning epistatic gene interactions from perturbation screens. 2021
 	  </a>
 	</font>
+
+	<aside class="notes">
+	  AF was published in 2018 (1), 2020 (2), 2024 (3).<br>
+	  Conclusion: Glinternet is in, xyz is out.
+	</aside>
 </section>
 
 <section>
@@ -132,7 +141,7 @@ <h4>Problem #2: Scalability</h4>
 
 <section>
 	<font size="6">
-	  Hierarchy and scalability are largely solved (until you sequence more)
+	  Hierarchy and scalability are largely solved (for now)
 	</font>
 	<img data-src="/assets/pint_summary.svg"><br>
 	<font size="3">
@@ -161,9 +170,13 @@ <h3>Model assumptions</h3>
 </section>
 
 <section>
-	<h4>Why don't you just do AI?</h4>
+	<h4>Why don't you just use AI?</h4>
 	<img data-src="/assets/linformer_model_selection.png" width=60%><br>
 	What AI?
+
+	<aside class="notes">
+	  Transformer 2017, Linformer 2020
+	</aside>
 </section>
 
 <section>
@@ -174,6 +187,10 @@ <h3>SNVformer</h3>
 		Kieran Elmes et al. SNVformer: an attention-based deep neural network for GWAS data. 2022
 	  </a>
 	</font>
+
+	<aside>
+	  Move technical details down under this slide for if we have time.
+	</aside>
 </section>
 
 <section>