To practice top-down design, start by working with a neighbor to design your main() function and sketch out all the functions that main() will need to call. Be sure to consider the types of all function parameters and return values, noting them in comments associated with each function.

After you have completed the design, implement the program and test it.

The handshake algorithm is an example of a linear algorithm, meaning that the number of steps required, which is ultimately related to the total run time of the algorithm, is directly proportional to the size of the data.

If we have a python list L of students (or student names, or integers, or whatever), where N=len(L), then, for a linear algorithm, the number of steps required will be proportional to N.

Computer scientists often write this using "big-O" notation as O(N), and we say the handshake algorithm is an O(N) algorithm (the O stands for "order", as in the "order of a function").

An O(N) algorithm is also called a linear algorithm because if you plot the number of steps versus the size of the data, you’ll get a straight line, like this:

Note on the plot that there are 3 straight lines. When comparing algorithms, you might have one algorithm that requires N steps, and another that requires 3*N steps. For example, if I modify the handshake algorithm to not only shake the student’s hand, but also ask them where they are from, and write down their email address, then this algorithm requires 3 steps (shake, ask, write down) for each student. So for N students, there would be 3*N total steps. Clearly the algorithm that requires the least steps would be better, but both of these algorithms are classified as "linear" or O(N). When classifying algorithms and using the big-O notation, we will ignore any leading constants. So algorithms with 3*N steps, N steps, N/2 steps, and 100*N steps are all considered as "linear".

For the introduction algorithm discussed above, how many introductions are required? We can count them and look for a pattern. For example, suppose I had only 6 students, then the number of introductions would be:

Remember, the first student is introduced to all of the others (5 intros), then the second student is introduced to all of the others, minus the first student, since they already met (4 intros), and so on.

If I had 7 students, then the number of introductions would be:

and if I had \$N\$ students:

\$(N-1) + (N-2) + (N-3) + ... + 3 + 2 + 1 =\$ ???

If you can see the pattern, awesome! Turns out the answer is \$\frac{(N-1)(N)}{2}\$ introductions for a class of size N. Try it for the N=6 and N=7 cases:

Another way to see that is with a bar chart showing how many introductions each student will make:

The bar chart just shows that student number 1 (on the x axis) has to do 33 introductions (the y axis), and student number 2 has to do 32, etc. And you can see how the bars fill up half of a 34x34 square. So if we had N students, the bar chart would fill up half of an NxN square.

For the equation above \$\frac{(N-1)(N)}{2}\$, if you multiply it out, the leading term is N-squared over 2 (i.e., half of an NxN square).

When comparing algorithms, we are mostly concerned with how they will perform when the size of the data is large. For the introductions algorithm, the number of steps depends directly on the square of N. This is written in big-O notation as \$O(N^2)\$ (order N squared). These types of algorithms are called quadratic, since the number of steps will quadruple if the size of the data is doubled.

Here’s a table showing the number of introductions versus the number of students in my class:

Notice how the number of introductions goes up very quickly as the size of the class is increased (compare to number of handshakes above).

For the RPS algorithm discussed above, how many times can I divide my class in two before I get down to just one student left?

To make things easier, assume I have a class of just 32 students. Then I can divide the students in two 5 times: 32→16→8→4→2→1.

We’re interested in knowing how the number of steps changes as the size of the data changes. If I double my class size to 64 students, how many more steps (divisions) will be required? The answer: just one extra division! That’s because the first division (64→32) gets us back to the class of just 32 students, and we know that only took 5 divisions, so a class of 64 will take 6 total divisions.

Again, if you can see the pattern here, awesome! If not, maybe this table will help:

Hopefully you can see that 2 raised to the 5th power is 32, and 2 to the 6th is 64 (try it in the python interactive shell: 2**5).

So for a class of N students, we want to know what x is in the equation 2**x=N. This is easily solved with a logarithm. Taking the base 2 log of N will get you x, which you can also see using the python interactive shell:

Finally, here’s a plot showing each type of algorithm we’ve looked at so far (plus one more we’ll see next week). All lines show number of steps (y axis) versus the size of the data (x axis).