Skip to content
Snippets Groups Projects
Commit 001f3262 authored by Ruth Anderson's avatar Ruth Anderson
Browse files

Update WriteUp.md

parent d6c1df04
No related branches found
No related tags found
No related merge requests found
......@@ -37,21 +37,22 @@ Throughout p1 and p2, you have written (or used) several distinct implementation
### BST vs. AVLTree ###
Construct input files for BST and AVLTree to demonstrate that an AVL Tree is asymptotically better
than a Binary Search Tree. To do this, we expect you to show trends. You might consider fitting a curve to
Construct inputs for BST and AVLTree to demonstrate that an AVL Tree is asymptotically better
than a Binary Search Tree. Comparing the worst case for each structure is fine here.
To do this, we expect you to show trends. You might consider fitting a curve to
your results. Explain your intuition on why your results are what they are.
<pre>TODO</pre>
### ChainingHashTable ###
Your ChainingHashTable should take as an argument to its constructor the type of "chains" it uses. Determine
which type of chain is (on average) best: an MTFList, a BST, or an AVL Tree. Explain your intuition on why
which type of chain is (on average, not worst case) best: an MTFList, a BST, or an AVL Tree. Explain your intuition on why
the answer you got makes sense (or doesn't!).
<pre>TODO</pre>
### Hash Functions ###
Write a new hash function (it doesn't have to be any good, but remember to include the code in your repository).
Compare the runtime of your ChainingHashTable when the hash function is varied. How big of a difference can the
hash function make? (You should keep all other inputs (e.g., the chain type) constant.) Explain your intuition on
hash function make (on average, not worst case)? (You should keep all other inputs (e.g., the chain type) constant.) Explain your intuition on
why your results are what they are.
<pre>TODO</pre>
......
0% Loading or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment