1.2. Errors in Numerical Methods

When using numerical methods, one must be aware of various types of errors that can affect the accuracy of solutions. These errors can arise from several sources, including approximations in calculations, the representation of numbers, and rounding errors.

This chunk introduces the concept of errors in numerical methods. It explains that when we apply numerical methods to solve problems, we are not guaranteed to get perfect answers. Instead, the answers can have inaccuracies, or errors, due to different factors like how calculations are approximated and how numbers are represented in computers.

Imagine following a recipe that requires precise measurements. If you measure ingredients with a spoon that has different sizes each time, your dish won't turn out the same way. Similarly, in numerical methods, small errors in representation or calculation can lead to incorrect or varied results.

1. Absolute Error:

• Absolute error is the difference between the exact value xexact and the approximate value xapprox.
• Formula: Absolute Error=∣xexact−xapprox∣
• Example: If the exact solution is 10.5 and the computed solution is 10.4, the absolute error is 0.1.

Absolute error is a straightforward way of measuring how far off an approximate solution is from the actual value. It's calculated by taking the absolute difference between the exact value and the approximate value. Absolute errors give us a sense of the raw difference but do not consider the size of the exact value, making it easy to understand but sometimes misleading.

Think about trying to hit a bullseye in archery. If you're aiming for 10 meters away and hit 9.9 meters, your absolute error is just 0.1 meters – but what if you were aiming from just 1 meter away? That same distance would feel much larger relative to your target. Absolute error, like that distance, tells you how far you are from your target but does not consider the overall scale.

2. Relative Error:

• Relative error is the absolute error normalized by the exact value. It gives an indication of the magnitude of error in relation to the size of the exact value.
• Formula: Relative Error=∣xexact−xapprox∣∣xexact∣
• Example: If the exact value is 10.5 and the computed value is 10.4, the relative error is 0.1/10.5≈0.0095.

Relative error provides context to the absolute error by expressing it as a fraction of the true value. This way, it shows how significant the error is compared to the actual value being measured. A small relative error, even if the absolute value seems large, may indicate a high level of accuracy if the actual value is also large.

Picture you're measuring the length of a football field (100 meters) with a tape that is 1 cm off. The absolute error is 1 cm, but when comparing that to the entire field, it becomes a tiny fraction (1 cm out of 100 m is negligible). In contrast, if you measure a small table (1 meter long) and are 1 cm off, that represents a 1% error, which is much more significant.

3. Round-off Error:

• Round-off errors occur because computers can only represent numbers to a limited number of decimal places or binary digits. This leads to small inaccuracies when representing numbers.
• For example, the number π cannot be exactly represented in a finite binary system, leading to round-off errors.

Round-off errors emerge because computers have a finite way of storing numbers, meaning they often have to cut off digits when representing complex numbers, leading to inaccuracy. This is particularly notable in irrational numbers like π, which cannot be precisely expressed in decimal or binary form.

Think of trying to read a clock. If it's 3.14159 hours and you round it to just 3.14, you've lost some precision. This tiny loss can accumulate in calculations, much like trying to gauge the long-distance running pace based on slightly off timing each lap.

4. Truncation Error:

• Truncation errors arise when an infinite process is approximated by a finite one. For example, in numerical integration, when using methods like the trapezoidal rule or Simpson’s rule, higher-order terms are truncated, leading to errors.
• In iterative methods, truncation errors occur because the exact solution is approached through a finite number of iterations.

Truncation error results from cutting off parts of a process that could theoretically continue indefinitely. When calculating areas or volumes, we might use simpler shapes to approximate more complex ones, leading to unavoidable inaccuracies. Similarly, in iterative methods, if we stop before reaching the final answer, we will have a truncation error because we haven’t fully solved the problem.

Imagine you're filling a bath. If you stop filling before it’s full, you might think it’s okay, but you don’t realize it’s not the exact level you wanted. Truncation errors are like leaving that tub partially filled when you actually wanted it completely full, leading to smaller inaccuracies in your measurements.

5. Algorithmic Error:

• These are errors introduced by the choice of numerical algorithm itself, such as in cases where an algorithm converges slowly or does not converge at all for certain inputs.

Algorithmic errors arise from the methods we select to solve problems. Some algorithms are better suited for certain situations and can converge to the correct answer quickly, while others may fail altogether or take too long to reach an acceptable solution. Choosing the right algorithm is essential to minimizing these errors.

Imagine trying to find your way to a new city using different routes: one may be direct and quick, while another involves many detours, making the trip longer. Selecting the right algorithm is similar to choosing the best route to reach your destination efficiently.

--