diff options
| author | Juan Marín Noguera <juan@mnpi.eu> | 2024-12-01 21:45:24 +0100 |
|---|---|---|
| committer | Juan Marín Noguera <juan@mnpi.eu> | 2024-12-01 21:45:24 +0100 |
| commit | e6a3f4af305b643955ce806a9614e65b91ef6325 (patch) | |
| tree | 2a7f177ce5c7058d5f3e629de05aafff6993aecd | |
| parent | 6b9ef86f54aa3acab4c55118382e6a98b006b301 (diff) | |
Exercises in 2.3.4.5 Path length
| -rw-r--r-- | 2.3.4.5.lyx | 332 | ||||
| -rw-r--r-- | index.lyx | 12 |
2 files changed, 344 insertions, 0 deletions
diff --git a/2.3.4.5.lyx b/2.3.4.5.lyx index e9a694e..8bd484a 100644 --- a/2.3.4.5.lyx +++ b/2.3.4.5.lyx @@ -107,5 +107,337 @@ TODO 3, \end_layout +\begin_layout Standard +\begin_inset ERT +status open + +\begin_layout Plain Layout + + +\backslash +rexerc3[M24] +\end_layout + +\end_inset + +An extended binary tree with +\begin_inset Formula $m$ +\end_inset + + external nodes determines a set of path lengths +\begin_inset Formula $l_{1},l_{2},\dots,l_{m}$ +\end_inset + + that describe the lengths of paths from the root to the respective external nodes. + Conversely, + if we are given a set of numbers +\begin_inset Formula $l_{1},l_{2},\dots,l_{m}$ +\end_inset + +, + is it always possible to construct an extended binary tree in which these numbers are the path lengths in some order? + Show that this is possible if and only if +\begin_inset Formula $\sum_{j=1}^{m}2^{-l_{j}}=1$ +\end_inset + +. +\end_layout + +\begin_layout Standard +\begin_inset ERT +status open + +\begin_layout Plain Layout + + +\backslash +answer +\end_layout + +\end_inset + +We assign each external node an interval of real numbers as follows: + start with +\begin_inset Formula $[l,u)=[0,1)$ +\end_inset + +; + then, + for each edge in the path from the root to the node, + if it's a left edge, + set +\begin_inset Formula $[l,u)\gets[l,\frac{l+u}{2})$ +\end_inset + +, + and if it's a right edge, + set +\begin_inset Formula $[l,u)\gets[\frac{l+u}{2},u)$ +\end_inset + +, + so the interval's length is +\begin_inset Formula $2^{-l}$ +\end_inset + +, + +\begin_inset Formula $l$ +\end_inset + + being the path length. + It's easy to see that these intervals are disjoint and that, + for each +\begin_inset Formula $x\in[0,1)$ +\end_inset + +, + there is a special node whose interval contains +\begin_inset Formula $x$ +\end_inset + +. + With this in mind: +\end_layout + +\begin_layout Itemize +\begin_inset Argument item:1 +status open + +\begin_layout Plain Layout +\begin_inset Formula $\implies]$ +\end_inset + + +\end_layout + +\end_inset + +This is precisely the sum of the lengths of the intervals, + which is 1 because +\begin_inset Formula $[0,1)$ +\end_inset + + is the disjoint union of these intervals. +\end_layout + +\begin_layout Itemize +\begin_inset Argument item:1 +status open + +\begin_layout Plain Layout +\begin_inset Formula $\impliedby]$ +\end_inset + + +\end_layout + +\end_inset + +We just have to sort the path lengths in increasing order, + assign consecutive intervals of length +\begin_inset Formula $2^{-l_{k}}$ +\end_inset + + starting from 0 and converting the intervals to paths (more precisely, + to sequences of left/right turns), + which we can do since the increasing order ensures that the starting point +\begin_inset Formula $l_{k}$ +\end_inset + + of an interval +\begin_inset Formula $[l_{k},u_{k})$ +\end_inset + + is a multiple of the length +\begin_inset Formula $u_{k}-l_{k}$ +\end_inset + +. + These paths are all different and none is a prefix of another one, + so they define the leaves of a binary tree. +\end_layout + +\begin_layout Standard +\begin_inset ERT +status open + +\begin_layout Plain Layout + + +\backslash +rexerc4[M25] +\end_layout + +\end_inset + +(E. + S. + Schwartz and B. + Kallick.) Assume that +\begin_inset Formula $w_{1}\leq w_{2}\leq\dots\leq w_{m}$ +\end_inset + +. + Show that there is an extended binary tree that minimizes +\begin_inset Formula $\sum w_{j}l_{j}$ +\end_inset + + and for which the terminal nodes in left to right order contain the respective values +\begin_inset Formula $w_{1},w_{2},\dots,w_{m}$ +\end_inset + +. +\end_layout + +\begin_layout Standard +\begin_inset ERT +status open + +\begin_layout Plain Layout + + +\backslash +answer +\end_layout + +\end_inset + +Let +\begin_inset Formula $T$ +\end_inset + + be an extended binary tree that minimizes +\begin_inset Formula $\sum w_{j}l_{j}$ +\end_inset + +. + For +\begin_inset Formula $i<j$ +\end_inset + +, + if +\begin_inset Formula $w_{i}<w_{j}$ +\end_inset + +, + then +\begin_inset Formula $l_{i}\geq l_{j}$ +\end_inset + + in that tree, + as otherwise we could reduce the weight path length by swapping their positions as +\begin_inset Formula +\[ +(w_{i}l_{j}+w_{j}l_{i})-(w_{i}l_{i}+w_{j}l_{j})=(w_{i}-w_{j})(l_{j}-l_{i})<0\#. +\] + +\end_inset + +Thus we might assume +\begin_inset Formula $l_{i}\geq l_{j}$ +\end_inset + + for all +\begin_inset Formula $i<j$ +\end_inset + +. + This means lengths +\begin_inset Formula $l_{1},\dots,l_{m}$ +\end_inset + + are in decreasing order, + so the proof in the previous exercise gives us a tree whose nodes, + in the order of the tree, + are +\begin_inset Formula $w_{m},\dots,w_{1}$ +\end_inset + + with lengths +\begin_inset Formula $l_{m},\dots,l_{1}$ +\end_inset + +, + and we just have to swap left and right edges in that tree. +\end_layout + +\begin_layout Standard +\begin_inset ERT +status open + +\begin_layout Plain Layout + + +\backslash +rexerc12[M20] +\end_layout + +\end_inset + +Suppose that a node has been chosen at random in a binary tree, + with each node equally likely. + Show that the average size of the subtree rooted at that node is related to the path length of the tree. +\end_layout + +\begin_layout Standard +\begin_inset ERT +status open + +\begin_layout Plain Layout + + +\backslash +answer +\end_layout + +\end_inset + +Let +\begin_inset Formula $V_{1},\dots,V_{m}$ +\end_inset + + be the internal nodes, + with path lengths +\begin_inset Formula $l_{1},\dots,l_{m}$ +\end_inset + +. + The average size of the subtrees is the sum of the number of nodes in each subtree divided by +\begin_inset Formula $m$ +\end_inset + +, + but in this sum, + each node +\begin_inset Formula $V_{k}$ +\end_inset + + is counted +\begin_inset Formula $l_{k}+1$ +\end_inset + + times, + one for the subtree generated by each node in the path from the root to +\begin_inset Formula $V_{k}$ +\end_inset + + including both ends of the path. + Thus this average is precisely +\begin_inset Formula +\[ +\frac{1}{m}\sum_{k}(l_{k}+1)=\frac{1}{m}(I+m)=\frac{I}{m}+1, +\] + +\end_inset + +where +\begin_inset Formula $I$ +\end_inset + + is the internal path length of the tree. +\end_layout + \end_body \end_document @@ -1689,6 +1689,18 @@ literal "false" \end_inset +\begin_inset Note Note +status open + +\begin_layout Plain Layout + +\family typewriter +A10+R25 +\end_layout + +\end_inset + + \end_layout \begin_layout Subsubsection |
