Sorting Algorithms and Analysis Robert Duncan
Refresher on Big-O O(2^N)Exponential O(N^2)Quadratic O(N log N)Linear/Log O(N)Linear O(log N)Log O(1)Constant Hierarchy of Big-O functions from slowest to fastest
Generic running times NO(log N)O(N)O(N log N)O(N^2)O(2^N) x10^ x10^308
O(N log N) vs. O(N^2)
Two Common Categories Sorting Algorithms of O(N^2) Bubble Sort Selection Sort Insertion Sort Sorting Algorithms of O(N log N) Heap Sort Merge Sort Quick Sort
For small values of N It is important to note that all algorithms appear to run equally as fast for small values of N. For values of N from the thousands to the millions, The differences between O(N^2) and O(N log N) become dramatically apparent
O(N^2) Sorts Easy to program Simple to understand Very slow, especially for large values of N Almost never used in professional software
Bubble Sort The most inefficient of the O(n^2) algorithms Simplest sorting algorithm available Works by comparing sequential items, and swapping them if the first one is larger than the second. It makes as many passes through an array as are needed to complete the sort
Bubble Sort – Pass
Bubble Sort – Pass
Bubble Sort – Pass
Bubble Sort – Pass
Selection Sort More efficient than Bubble Sort, but not as efficient as Insertion Sort Works by finding the largest element in the list and swapping it with the last element, effectively reducing the size of the list by 1.
Selection Sort – Pass
Selection Sort – Pass
Insertion Sort One of the most efficient of the O(n^2) algorithms Roughly twice as fast as bubble sort Works by taking items from unsorted list and inserting them into the proper place.
Insertion Sort
Insertion Sort
O(N log N) Sorts Fast Efficient Complicated, not easy to understand Most make extensive use of recursion and complex data structures
Heap Sort Slowest O(N log N) algorithm. Although the slowest of the O(N log N) algorithms, it has less memory demands than Merge and Quick sort.
Heap Sort Works by transferring items to a heap, which is basically a binary tree in which all parent nodes have greater values than their child nodes. The root of the tree, which is the largest item, is transferred to a new array and then the heap is reformed. The process is repeated until the sort is complete.
Forming the heap from an unsorted array
Populating the new array
Reforming the heap
Reforming the heap
Repeat the process
Repeat the process
Repeat the process
Repeat the process
Merge Sort Uses recursion. Slightly faster than heap, but uses twice as much memory from the 2 nd array. Sometimes called “divide and conquer” sort. Works by recursively splitting an array into two equal halves, sorting the items, then re- merging them back into a new array.
Quick Sort The most efficient O(N log N) algorithm available Works by first randomly choosing a pivot point which is hopefully a median element in the array. All elements less than the pivot are transferred to a new array, and all elements greater are transferred to a second array.
Quick Sort These new arrays are recursively quick sorted until they have been split down to a single element. The sorted elements smaller than the pivot are placed in a new array, then the pivot is placed after, and finally the elements greater than the pivot. These elements are now sorted.
Quick Sort Note: Although it is the quickest sorting algorithm, a badly chosen pivot may cause Quick Sort to run at O(N^2). Different versions of quick sort address this problem in different ways. For example, one way is to randomly choose 3 elements, then use the median element as the pivot.
Pivot
Pivot
11275 Pivot
Pivot
Pivot
What’s the point
Binary Searching Binary searches achieve O(log N) efficiency on sorted data Similar to High-Low game Each execution eliminates half of the elements to search for Although hashing offers a quicker search of O(1), binary searches are simpler, and use much less memory.
Binary Searching ?
Binary Searching ?
Binary Searching
Conclusion… O(N^2) algorithms are almost never used in professional software Quick sort is generally considered to be the best overall sorting algorithm currently available.