Class Notes (809,756)
Canada (493,780)
CMPT 225 (60)
John Edgar (28)
Lecture 4

CMPT 225 Week 4 Lecture 3

2 Pages
Unlock Document

Simon Fraser University
Computing Science
CMPT 225
John Edgar

Best Case,Average Case, Worst Case How performance varies for different input of same size – based on organization of the data. May be asked what an array should look like to be best/avg/worst case for an algorithm. Although useful, can be difficult to determine exact num of operations by an algorithm Alternative is to focus on barometer instruction Is instruction executed most times in an algorithm. May be multiple. Number of times this is executed is usually proportional to the running time of the algorithm. Cost Functions for Searching If array isn't sorted, use linear search (it's the only way to do it) - start with first item, compare each item to target, if find, return true, else never find, return false Linear search barometer instruction: equality checking (comparisons) arr[i] == x. Or, could be i++, or I < arr.length. Best case: 1, item looking for is first item in array → make comparison arr[i] == x once. Worst case: length of the array, or not in the array. Both cases: have to go through entire thing. Make n comparisons. Average case: to calc, need to know how often the two worst cases arise. Can make assumptions about this, although they might not hold for a particular use. Essentially, how often are you going to look for something in the array? How often will you find something in the end of the array? How often at the start? Assume: target is not in array half the time. Therefor half the time entire array has to be checked. Probability 0.5. Also assume: Equal probability of target being at any array location. Probability 1/n that target is at some position i. I comparisons if target at the ith position. Take weighted average of values to find the total expected number of comparisons. E = 1*1/n + 2*1/n + … + n * 1/n or E = (n+1)/2 Target not in array: n comparisons Target in array: (n+1)/2 comparisons Take weighted average = (3n + 1)/4 So, on average, we assume it will perform this many comparisons. If we sorted, could change average cost to n/2. But then we shouldn't be using
More Less

Related notes for CMPT 225

Log In


Don't have an account?

Join OneClass

Access over 10 million pages of study
documents for 1.3 million courses.

Sign up

Join to view


By registering, I agree to the Terms and Privacy Policies
Already have an account?
Just a few more details

So we can recommend you notes for your school.

Reset Password

Please enter below the email address you registered with and we will send you a link to reset your password.

Add your courses

Get notes from the top students in your class.