Held: Monday, 23 February 2004
Summary:
Today we consider techniques for comparatively evaluating the running
time of algorithms.
Related Pages:
Notes:
- Some of you asked about the relationship between tail recursion and iteration. We'll discuss it in class today.
- Are there questions on homework 4?
- An optional reading on higher-order procedures should be available outside my office this afternoon.
Overview:
- Comparing Algorithms.
- Asymptotic Analysis.
- Eliminating Constants.
- Asymptotic Analysis in Practice.
- The Role of Details.
- As you may have noted, there are often multiple algorithms one can
use to solve the same problem.
- In finding the minimum element of a list, you can step through
the list, keeping track of the current minimum. You could also
sort the list and grab the first element.
- In finding x^{y}, one might use repeated multiplication,
divide and conquer, or even the built-in e^{n} and
natural log procedures.
- You can come up with your own variants.
- How do we choose which algorithm is the best?
- The fastest/most efficient algorithm.
- The one that uses the fewest resources.
- The clearest.
- The shortest.
- The easiest to write.
- The most general.
- ...
- Frequently, we look at the
speed
. That is, we consider how
long the algorithm takes to run.
- It is therefore important for us to be able to analyze the running
time of our algorithms.
- Is there an exact number we can provide for the running
time of an algorithm?
- Different inputs lead to different running times. For example,
if there are conditionals in the algorithm (as there are in many
algorithms), different instructions will be executed depending
on the input.
- Not all operations take the same time. For example, addition is
typically quicker than multiplication, and integer addition is
typically quicker than floating point addition.
- The same operation make take different times on different machines.
- The same operation may appear to take different times on the same machine,
particularly if other things are happening on the same machine.
- Many things that affect running time happen behind the scenes and
cannot be easily predicted. For example, the computer might move
some frequently-used data to cache memory.
- Noting problems in providing a precise analysis of the running time of
programs, computer scientists developed a technique which
is often called asymptotic analysis. In asymptotic analysis of
algorithms, one describes the general behavior of algorithms in terms of
the size of input, but without delving into precise details.
- The analysis is
asymptotic
in that we look at the behavior
as the input gets larger.
- There are many issues to consider in analyzing the asymptotic behavior
of a program. One particularly useful metric is an upper bound
on the running time of an algorithm. We call this the
Big-O
of
an algorithm.
- Big-O is defined somewhat mathematically, as a relationship between
functions.
- f(n) is in O(g(n)) iff
- there exists a number n_{0}
- there exists a number d > 0
- for all n > n_{0},
abs(f(n)) <= abs(d*g(n))
- What does this say? It says that after a certain value,
n_{0},
f(n) is bounded above by a constant (that is, d)
times g(n).
- The constant, d, helps accommodate the variation in the algorithm.
- We don't usually identify the d precisely.
- The n_{0} says
for big enough n
.
- We can apply big-O to algorithms.
- n is the
size
of the input (e.g., the number of items in
a list or vector to be manipulated).
- f(n) is the running time of the algorithm.
- Some common Big-O bounds
- An algorithm that is in O(1) takes constant time. That is, the
running time is independent of the input. Getting the size of a
vector should be an O(1) algorithm.
- An algorithm that is in O(n) takes time linear in the size of
the input. That is, we basically do constant work for each
element
of
the input. Finding the smallest element in a list is often an
O(n) algorithm.
- An algorithm that is in O(log_{2}(n)) takes logarithmic time.
While the running time is dependent on the size of the input,
it is clear that not every element of the input is processed.
Many such algorithms involve the strategy of
divide and conquer
.
- One of the nice things about asymptotic analysis is that it makes
constants
unimportant
because they can be hidden
in the d.
- If f(n) is 100*n seconds and g(n)
is 0.5*n seconds, then
- f(n) is in O(g(n))
[let d be 200]
- g(n) is in f(n)
[let d be 1]
- If f(n) is 100*n seconds and g(n) is
n*n seconds, then
f(n) is in O(g(n))
[let n_{0} be 100 and d be 1]
- However, g(n) is not in O(f(n)).
Why not?
- Suppose there were an n_{0} and a d.
- Consider what happens for n = 101*d.
- d*f(n) = d*100*101*d =
d*d*100*101.
- However, g(n) = d*d*101*101, which is
even larger.
- If n_{0} is greater than 101d, we'll still have
this problem [proof left to reader].
- Since constants can be eliminated, we normally don't write them.
- That is, we say that the running time of an algorithm is
O(n) or O(n^{2}) or ....
- We now have a theoretical grounding for asymptotic analysis. How
do we do it in practice?
- At this point in your career, it's often best to
count
the steps
in an algorithm and then add them up. After you've taken combinatorics,
you can use recurrence relations.
- Over the next few days, we'll look at a number of examples. Some
starting ones.
- Finding the smallest/largest element in a vector of integers.
- Finding the average of all the elements in a vector of integers.
- Putting the largest element in an array at the end of the array.
if we're only allowed to swap subsequent elements.
- Computing the nth Fibonacci number.
- ...
- Although the previous discussion may make it seem like precise
details of algorithms aren't important, some details are
particularly important.