You are on page 1of 6

1

Again and again


A common technique to nd new and interesting patterns in mathematics is iterationsimply put, do it again. Here are some examples. 1. Binary searches Suppose someone picks a number from 1 to 100. You get to guess the number, and the person will tell you if youre too high or too low. What is the maximum number of guesses you will ever need? A very ecient way to get the answer is to use a binary search. Youve probably heard that the prex bi means two. In a binary search, you cut the possible answers into two equal (or nearly equal) groups. Discard the half that you know doesnt include what youre looking for, then repeat the process with what remains. (a) For the number search, try going for the middle number the arithmetic mean of the highest possible and lowest possible. (Round down when needed.) For example, for the rst guess, the lowest possible is 1 and the highest possible is 100. The arithmetic mean is 100+1 2 or 50.5, so guess 50. If this is too low, the lowest possible is 51 and the highest possible is 100. . . . Iterate the process: the next guess should be the arithmetic mean of 51 and 100 (rounded down). Try to nd the largest number of guesses required to nd a number from 1 to 100 this way. (b) Binary searches can also be used by computer programs to nd names, or to alphabetize a list of words. Suppose a program were searching the following list of names (sorted in alphabetical order): Abraham, Elizabeth; Berkowitz, Carl; Docherty, Marian; Evans, Revonda; Gonzales, Carlos; Jones, Vivian; Lloyd, Fiona; Nguyen, John; Smith, Kara; Young, Mel; Ze, John i. If the program compared each name in the order given, what is the least number of comparisons it would have to make to nd a name? (For example, to nd Marian Docherty, it would rst see that Abraham is not Docherty, so it would then move to the next name, Berkowitz.) What is the greatest number of comparisons it would have to make?
c EDC 2001

Problems with a Point: August 29, 2001

Again and again: Problem

ii.

For a binary search, the program would pick the middle name and compare it to the name its looking for. If the name is identical, the program found it. If the name picked comes before the name its looking for (alphabetically), it would focus on just the second half of the names. Similarly, if the name picked comes after the name its looking for, it would focus on just the rst half of the names. Using this strategy, whats the least number of comparisons it would have to make? What is the greatest number?

2.

Geometry and fractals Imagine this: Start with a line segment. Then, on each side, erase the middle third and add two segments the same length (as if an equilateral triangle had grown o it). Iterate this with each segment of the new gure: erase the middle third and add two new segments of the same size.

If you imagine iterating forever, you get a gure that is a fractal. (If you start with an equilateral triangle instead of just a segment, the result is often called a Koch Snowake.) One feature of most fractals is self-similarity, that is, if you magnify a part of the gure, it will look just like the gure (or a portion of the gure) before you magnied it. Heres another, sometimes called Sierpinskis Carpet. On graph paper, draw a square with 81 small squares on each side. (If you cant get 81 squares on each side, use 27 per side instead.) Inside the square, make 9 smaller squares and shade the middle one. (See the diagram on the next page.) Iterate the process with each of the smaller unshaded squares: make 9 smaller squares and shade the middle one. Do this as much as you can.
Problems with a Point: August 29, 2001 c EDC 2001

Stage 5 Koch Snowake:

Then imagine doing it forever!

Again and again: Problem

3.

Numerical patterns Iteration appears in numerical patterns, too. (a) For the following pattern from problem 2, pretend the initial segment has a length of 1 foot.

Find the total length in the rst several stages. Find a numerical way to describe this pattern? (b) For the Sierpinskis Carpet, assume the sides of the initial square were each 1 foot long. Find the area of the unshaded part of the rst several stages. Describe the pattern. Can you predict the area at a given stage? (c) For the next pattern, use this Pascals Triangle:

Problems with a Point: August 29, 2001

c EDC 2001

Again and again: Problem

Count how many odd numbers there are in each row, and nd a pattern in the results. (Youll probably want to look at 16 rows at least!) (d) Binary numbers have only two digits, 0 and 1. The rst whole numbers (0, 1, 2, 3, 4, . . . ) are then 0, 1, 10, 11, 100, 101, 111, 1000, and so on. To go from the current number to the next, change the right-most digit. If its a 0, change it to 1 and you have the next number. If its a 1, change it to 0 and then change the digit to the left. (This is the same thing you do with our usual system: If the right-most number is a 9, change it to 0 and then add 1 to the digit to the left. If that digits a 9. . . .) Write the numbers 016 in binary form (only 0s and 1s). Then nd how many 1s appear in each number. How is the pattern in the number of 1s like the pattern in problem 3c?

This is one wayan iterative wayto nd the next binary number.

Problems with a Point: August 29, 2001

c EDC 2001

Again and again: Answers

Answers
1. (a) The largest number of guesses required is 7. (b) i. The least number of comparisons is 1; the greatest is 11. ii. The least number of comparisons is 1; the greatest is 4. 2. Counting the 81-by-81 square (no shaded parts) as Stage 1, students can get to Stage 5 without reducing the graph papers squares further:

3. (a)

The lengths in feet are 1, 4 , 16 , 64 , . . . . To get the length 3 9 27 for the next stage, multiply by 4 . That is, the length 3
n1

at stage n is 4 . 3 (b) The areas in square feet are 1, 8 , 64 , 512 , . . . . That is, 9 81 729 the area at stage n is . (c) The number of odd numbers per row in this triangle are 1, 2, 2, 4, 2, 4, 4, 8, 2, 4, 4, 8, 4, 8, 8, and 16. At this point, how the pattern will continue may still not be clear, but guesses that correctly apply to these rst 16 rows should be accepted. One way to describe the true pattern is to start with 1. Go through the previous pattern from the beginning, doubling the number. When you get to end of the previous pattern (and have now doubled the number of rows), start at the beginProblems with a Point: August 29, 2001 c EDC 2001

8 9

n1

Teachers Note: You might have students calculate this area for stages late in the process, such as stage 1000 or higher. Calculators will probably report the area as 0 at stage 1936 and higher. Students may nd it surprising that there are clearly points that will never be removed (such as the sides of the initial square) but the area approaches 0 when the process is iterated indenitely.

Again and again: Answers

ning again. That is, start with 1; then start over, doubling: 2. You now have 1, 2. Start over and double: 2, 4. You now have 1, 2, 2, 4. Start over and double: 2, 4, 4, 8. You now have 1, 2, 2, 4, 2, 4, 4, 8. And so on. (d) The rst sixteen numbers and the number of 1s are as follows: Decimal Binary Number number number of 1s 0 0 0 1 1 1 2 10 1 3 11 2 4 100 1 5 101 2 6 110 2 7 111 3 8 1000 1 9 1001 2 10 1010 2 11 1011 3 12 1100 2 13 1101 3 14 1110 3 15 1111 4 16 10000 1 This is just like the Pascals Triangle pattern, except you start with 0 instead of 1 and you add 1 instead of double: 0, then add one: 1. You now have 0, 1. Add one: 1, 2. You now have 0, 1, 1, 2. And so on.

Problems with a Point: August 29, 2001

c EDC 2001

You might also like