general refactoring

Makefile

@@ -2,18 +2,17 @@ ENV_NAME = rlbook
 
 RUN = micromamba run -n $(ENV_NAME)
 
-_NOTEBOOKS = $(addprefix book/, intro bandits mdps fitted_dp control pg exploration)
-
-NOTEBOOKS = $(addsuffix .md, $(_NOTEBOOKS))
-
-IPYNBS = $(addsuffix .ipynb, $(_NOTEBOOKS))
+_NOTEBOOKS = $(addprefix book/, bandits contextual_bandits control exploration fitted_dp imitation_learning mdps pg planning supervised_learning)
 
 _META = \
-	appendix \
+	background \
 	bibliography \
-	challenges \
 	index
 
+NOTEBOOKS = $(addsuffix .md, $(_NOTEBOOKS))
+
+IPYNBS = $(addsuffix .ipynb, $(_NOTEBOOKS))
+
 META = $(addsuffix .md, $(addprefix book/, $(_META)))
 
 SOLUTIONS = book/solutions/bandits.py
@@ -50,3 +49,6 @@ lab:
 
 lint:
 	$(RUN) ruff check --fix $(IPYNBS)
+
+publish: book/_build/html
+	$(RUN) ghp-import --cname "rlbook.adzc.ai" --no-jekyll --push --force book/_build/html

book/_toc.yml

@@ -6,14 +6,15 @@ root: index.md
 options:
   numbered: true
 chapters:
-  - file: intro.md
-  - file: bandits.md
   - file: mdps.md
-  - file: fitted_dp.md
   - file: control.md
+  - file: bandits.md
+  - file: supervised_learning.md
+  - file: fitted_dp.md
   - file: pg.md
-  - file: exploration.md
   - file: imitation_learning.md
-# - file: challenges
-# - file: appendix
+  - file: planning.md
+  - file: exploration.md
+  - file: contextual_bandits.md
   - file: bibliography.md
+  - file: background.md

book/background.md

@@ -0,0 +1,37 @@
+(background)=
+# Appendix: Background
+
+## O notation
+
+Throughout this chapter and the rest of the book, we will describe the
+asymptotic behavior of a function using $O$ notation.
+
+For two functions $f(t)$ and $g(t)$, we say that $f(t) \le O(g(t))$ if
+$f$ is asymptotically upper bounded by $g$. Formally, this means that
+there exists some constant $C > 0$ such that $f(t) \le C \cdot g(t)$ for
+all $t$ past some point $t_0$.
+
+We say $f(t) < o(g(t))$ if asymptotically $f$ grows strictly slower than
+$g$. Formally, this means that for *any* scalar $C > 0$, there exists
+some $t_0$ such that $f(t) \le C \cdot g(t)$ for all $t > t_0$.
+Equivalently, we say $f(t) < o(g(t))$ if
+$\lim_{t \to \infty} f(t)/g(t) = 0$.
+
+$f(t) = \Theta(g(t))$ means that $f$ and $g$ grow at the same rate
+asymptotically. That is, $f(t) \le O(g(t))$ and $g(t) \le O(f(t))$.
+
+Finally, we use $f(t) \ge \Omega(g(t))$ to mean that $g(t) \le O(f(t))$,
+and $f(t) > \omega(g(t))$ to mean that $g(t) < o(f(t))$.
+
+We also use the notation $\tilde O(g(t))$ to hide logarithmic factors.
+That is, $f(t) = \tilde O(g(t))$ if there exists some constant $C$ such
+that $f(t) \le C \cdot g(t) \cdot \log^k(t)$ for some $k$ and all $t$.
+
+Occasionally, we will also use $O(f(t))$ (or one of the other symbols)
+as shorthand to manipulate function classes. For example, we might write
+$O(f(t)) + O(g(t)) = O(f(t) + g(t))$ to mean that the sum of two
+functions in $O(f(t))$ and $O(g(t))$ is in $O(f(t) + g(t))$.
+
+## Python
+
+