Recompile lectures with IEI-style compile.sh and shared diagrams

_lamd/_lamd.yml

@@ -13,15 +13,15 @@ url: https://inverseprobability.com/
 talkcss: https://inverseprobability.com/assets/css/talks.css
 bibdir: $HOME/lawrennd/bibliography
 postsdir: ../_lectures/
-slidesdir: $HOME/lawrennd/slides/
+slidesdir: ../slides/ # Where compiled slides go
 notesdir: ../_notes/
 notebooksdir: ../_notebooks/
 practicalsdir: ../_practicals/
 snippetsdir: $HOME/lawrennd/snippets
 macrosdir: $HOME/lawrennd/lamd/lamd/macros
 writediagramsdir: .
 scriptsdir: https://inverseprobability.com/talks/scripts/
-diagramsdir: ../slides/diagrams/
+diagramsdir: $HOME/lawrennd/slides/diagrams/ # Shared figure library (copy_web_diagrams reads from here)
 slidesheader: slides-header.html
 notation: talk-notation.tex
 posts: True

_lamd/bayesian-regression.md

@@ -83,20 +83,31 @@ $$}
 
 \addreading{@Bishop:book06}{Section 1.2.3 (pg 21–24)}
 \addreading{@Rogers:book11}{Sections 3.1-3.4 (pg 95-117)}
-\addreading{@Bishop:book06}{Section 1.2.3 (pg 21–24)}
 \addreading{@Bishop:book06}{Section 1.2.6 (start from just past eq 1.64 pg 30-32)}
 
 \reading
 
 \include{_physics/includes/gauss-least-squares.md}
-\include{_ml/includes/the-bayesian-approach.md}
-\include{_ml/includes/bayesian-regression1d.md}
-\include{_ml/includes/bayesian-1d-maths.md}
+\include{_ml/includes/bayesian-regression1d-short.md}
+\include{_ml/includes/bayesian-regression1d-maths.md}
+
+\notes{
+\subsection{Bayesian Inference in the Univariate Case}
+
+This video talks about Bayesian inference across the single parameter, the offset $c$, illustrating how the prior and the likelihood combine in one dimension to form a posterior.
+
+\figure{\includeyoutube{AvlnFnvFw_0}{600}{450}{15}}{Univariate Bayesian inference. Lecture 10 from 2012 MLAI Course.}{univariate-bayesian-inference-video}
+}
 
 \subsection{The Joint Density}
 
+\slides{
 * Really want to know the *joint* posterior density over the parameters $c$ *and* $m$.
-* Could now integrate out over $m$, but it’s easier to consider the multivariate case.
+* Could now consider $m$, but it is easier to consider the multivariate case.
+}
+\notes{This gives us the distribution over $c$, but in reality we want to know the *joint* posterior density over the parameters $c$ and $m$. We could find this by now considering $m$, but it is easier to consider the multivariate case.}
+
+\include{_ml/includes/the-bayesian-approach.md}
 
 \include{_ml/includes/two-d-gaussian.md}
 
@@ -147,7 +158,7 @@ sigma2 = 0.01}
 * Revisit maximum likelihood solution with validation set.
 }
 
-\newslide{Regularized Mean}
+\newslide{Regularised Mean}
 
 \slides{
 * Validation fit here based on mean solution for $\mappingVector$ only.
@@ -164,6 +175,7 @@ sigma2 = 0.01}
 * In other cases $\alpha \eye$ *regularizes* the system (keeps parameters smaller).
 }
 \include{_ml/includes/posterior-sampling-basis.md}
+
 \include{_ml/includes/polynomial-marginal-likelihood.md}
 \include{_ml/includes/compute-output-expectations.md}
 

_lamd/compile.sh

@@ -1,11 +1,43 @@
-#/bin/bash
-
-FILES=""
-SKIP=true
-while read stub; do
-    if $SKIP; then
-	SKIP=false
-    else
-	maketalk $stub
-    fi
-done < lectures.csv
+#!/bin/bash
+# Compile lectures and worksheets listed in the CSVs (skip header row).
+set -euo pipefail
+
+cd "$(dirname "$0")"
+
+compile_list() {
+    local csv="$1"
+    local skip=true
+    local stub
+    while IFS= read -r stub || [[ -n "$stub" ]]; do
+        stub="${stub//$'\r'/}"
+        [[ -z "$stub" ]] && continue
+        if $skip; then
+            skip=false
+            continue
+        fi
+        echo "maketalk ${stub}.md"
+        maketalk "${stub}.md"
+    done < "$csv"
+}
+
+compile_list lectures.csv
+if [[ -f worksheets.csv ]]; then
+    compile_list worksheets.csv
+fi
+
+# Printable assessment handouts (pandoc), not LaMD.
+handouts_dir="../assessments/handouts"
+if [[ -f "${handouts_dir}/handouts.csv" ]]; then
+    skip=true
+    while IFS= read -r stub || [[ -n "$stub" ]]; do
+        stub="${stub//$'\r'/}"
+        [[ -z "$stub" ]] && continue
+        if $skip; then
+            skip=false
+            continue
+        fi
+        echo "pandoc ${stub}.md -> tex/pdf"
+        pandoc "${handouts_dir}/${stub}.md" -s -o "${handouts_dir}/${stub}.tex"
+        pandoc "${handouts_dir}/${stub}.md" -o "${handouts_dir}/${stub}.pdf"
+    done < "${handouts_dir}/handouts.csv"
+fi

_lamd/dimensionality-reduction.md

@@ -17,8 +17,8 @@ youtube: 0mtK2_rc0IY
 \notes{So far in our classes we have focussed mainly on regression
 problems, which are examples of supervised learning. We have considered the
 relationship between the likelihood and the objective function and we have shown
-how we can find paramters by maximizing the likelihood (equivalent to minimizing
-the objective function) and in the last session we saw how we can *marginalize*
+how we can find parameters by maximising the likelihood (equivalent to minimising
+the objective function) and in the last session we saw how we can *marginalise*
 the parameters in a process known as Bayesian inference.}
 
 
@@ -30,10 +30,14 @@ the parameters in a process known as Bayesian inference.}
 
 \include{_dimred/includes/probabilistic-pca.md}
 
-\include{_dimred/includes/mocap-ppca.md}
+\include{_dimred/includes/osu-run1-ppca.md}
 \include{_dimred/includes/robot-wireless-ppca.md}
+
 \include{_dimred/includes/ppca-interpretations.md}
-\include{_dimred/includes/pca-in-practice.md}
+\include{_dimred/includes/pca-effectiveness.md}
+
+\section{Derivation of PPCA}
+
 \include{_dimred/includes/ppca-marginal-likelihood.md}
 \include{_dimred/includes/ppca-reconstruction.md}
 

_lamd/linear-regression.md

@@ -16,14 +16,11 @@ youtube: 5VPr6NbHHjg
 * This time: explore least squares for regression.
 
 \include{_ml/includes/regression-examples.md}
-
-\include{_datasets/includes/olympic-100m-data.md}
-
-
 \include{_datasets/includes/olympic-marathon-data.md}
 \include{_ml/includes/what-is-ml.md}
 \include{_ml/includes/sum-of-squares-error.md}
 \include{_ml/includes/linear-algebra-regression.md}
+\include{_ml/includes/linear-regression-hessian.md}
 
 \thanks