UMBC CMSC202, Computer Science II, Spring 1998, Sections 0101, 0102, 0103, 0104 and Honors

Thursday February 17, 1998

Assigned Reading:

• A Book on C:
• Programming Abstractions in C:

Handouts (available on-line): Asymptotic Analysis

Topics Covered:

• Rehashed tweaking Merge Sort and Quicksort for better performance. (See previous lecture.) This only worked because insertion sort is "faster" for small input sizes.

• Sample runs on ordered and random data show that the running times for Insertion Sort and Quicksort varies greatly. Insertion Sort is fastest on an array that is already sorted and slower than Selection Sort if the data is sorted in the opposite order. Quicksort works well on random data, but is much slower when the data is sorted in either order. Selection Sort and Merge Sort on the other hand seem to run in approximately the same amount of time regardless of the ordering of the data. We used a new program to generate files of ordered data. (Random data was generated previously.)

• To distinguish Quicksort and Merge Sort (which are quick) from Selection Sort and Insertion Sort (which are slow), we use asymptotic analysis. (See online notes.) In this approach, we say that the worst case running times of Selection Sort and Insertion Sort are O(n²). The worst case running time of Merge Sort is O(n log n) and the average case running time of Quicksort is O(n log n). Note that the Big Oh notation ignores the running times of small inputs and also ignores constant factors. In this scheme, we lump Quicksort and Merge Sort together as "fast" O(n log n) algorithms and we lump Selection Sort and Insertion Sort together as "slow" O(n²) algorithms.

• Next, we considered the difference between the running times of linear search versus binary search. Our linear search program looks for 100,000 random integers in a sorted array of 800,000 random integers. (Note that the number of times we find the random number in the array corresponded to our predictions, which were based upon the number of positive 4-byte integers.) The binary search program does the same thing. However, the sample runs for linear search and for binary search show that binary search is much faster. In this case, we say that linear search is a O(n) algorithm, whereas binary search is a O(log n) algorithm.