Number Set Arithmetic and Precision

Types of Arithmetic Systems Introduction When we perform calculations, different number systems behave in different ways. Some arithmetic operations that work perfectly with whole numbers fail with fractions, and some operations that work with fractions become problematic when computers represent numbers with limited precision. Understanding these distinctions is essential because it explains why calculations sometimes produce unexpected results and how to account for errors in real-world computation. Integer Arithmetic Integers are whole numbers like −3, 0, 5, and 42. While integer arithmetic feels intuitive, it has an important limitation: it is not closed under division. This means dividing one integer by another doesn't always produce an integer. For example, $7 \div 2 = 3.5$, which is not an integer. When a programming language performs "integer division," it typically truncates (cuts off) the decimal part, so $7 \div 2$ would yield just $3$. This behavior differs from the division you might perform with a calculator. Rational Number Arithmetic A rational number is any number that can be expressed as a fraction $\frac{a}{b}$ where $a$ and $b$ are integers and $b \neq 0$. Rational arithmetic is where fractions become important. Adding and Subtracting Fractions When fractions share the same denominator, addition is straightforward: $$\frac{a}{c} + \frac{b}{c} = \frac{a+b}{c}$$ For example, $\frac{1}{4} + \frac{3}{4} = \frac{4}{4} = 1$. When denominators differ, you must first find a common denominator. Multiplying the first fraction by $\frac{d}{d}$ and the second by $\frac{c}{c}$ produces: $$\frac{a}{c} + \frac{b}{d} = \frac{ad}{cd} + \frac{bc}{cd} = \frac{ad + bc}{cd}$$ For example, $\frac{1}{2} + \frac{1}{3} = \frac{3}{6} + \frac{2}{6} = \frac{5}{6}$. Subtraction follows the same pattern with a minus sign between numerators: $\frac{a}{c} - \frac{b}{d} = \frac{ad - bc}{cd}$. Multiplying and Dividing Fractions Multiplication is simpler than addition: multiply the numerators together and the denominators together: $$\frac{a}{b} \times \frac{c}{d} = \frac{ac}{bd}$$ Division is performed by multiplying by the reciprocal (flipping the second fraction): $$\frac{a}{b} \div \frac{c}{d} = \frac{a}{b} \times \frac{d}{c} = \frac{ad}{bc}$$ Closure Under Operations A key property of rational arithmetic is that it is closed under addition, subtraction, multiplication, and division (except division by zero). This means performing any of these operations on two rational numbers always produces another rational number. For instance, $\frac{1}{2} + \frac{1}{3} = \frac{5}{6}$ is still rational. Decimal Representations of Fractions Every rational number can be expressed as a decimal. Decimal fractions have denominators that are powers of 10: $$0.75 = \frac{75}{100} = \frac{3}{4}$$ Some rational numbers produce repeating decimals, where a sequence of digits repeats infinitely: $$0.\overline{3} = 0.333... = \frac{1}{3}$$ The bar over the 3 indicates that the digit repeats. Similarly, $0.\overline{142857} = \frac{1}{7}$. Real Number Arithmetic Real numbers include all rational numbers plus irrational numbers like $\pi$ and $\sqrt{2}$, which cannot be expressed as fractions. Real number arithmetic introduces new challenges, particularly around precision. Truncation and Rounding When working with decimals, we often cannot keep all digits. Two common approaches are: Truncation discards all digits beyond a chosen position. For example, $\pi = 3.14159...$ truncated to three decimal places is $3.141$. Rounding adjusts the last retained digit based on what comes next. If the first discarded digit is 5 or greater, round up; otherwise, round down. The same value of $\pi$ rounded to three decimal places is $3.142$ (since the next digit, 1, causes us to round the final 1 up to 2). Floating-Point Representation Computers cannot store real numbers with infinite precision. Instead, they use floating-point representation, which expresses numbers in the form: $$\text{significand} \times \text{base}^{\text{exponent}}$$ For example, $1,250 = 1.25 \times 10^3$ in decimal floating-point, or $12 = 1.5 \times 2^3$ in binary floating-point. The number of bits allocated to store the significand and exponent determines the precision and range of representable numbers. Rounding Errors A critical limitation of floating-point arithmetic is that rounding errors occur whenever a result requires more bits than are available. The computed value is rounded to the nearest representable number, introducing error. These errors are small individually but can accumulate when many operations are performed. Approximation, Errors, and Significant Digits Scientific and engineering work relies on measurements that are inherently inexact. Managing these uncertainties is crucial for reporting results responsibly. What Are Significant Digits? Significant digits are the digits in a number that represent its precision and certainty. In a measurement, they indicate the resolution of the instrument and the confidence in the result. Zeros can be tricky: Leading zeros (like the 0 in 0.025) are never significant; they only indicate position. Trailing zeros without a decimal point (like the zeros in 1,200) are generally not significant. Trailing zeros after a decimal point (like the zeros in 1.200) are always significant. Zeros between non-zero digits are always significant. For example: $0.00456$ has 3 significant digits (4, 5, and 6) $1.05$ has 3 significant digits (1, 0, and 5) $3,000$ has 1 significant digit (3), but $3,000.0$ has 5 significant digits Propagation of Uncertainty When combining measurements through arithmetic: For addition and subtraction, add the absolute uncertainties. If you measure 15.2 cm (±0.1 cm) plus 8.7 cm (±0.1 cm), the uncertainty of the sum is ±0.2 cm. For multiplication and division, add the relative uncertainties (percentage uncertainties). If one value is known to within 2% and another to within 3%, the product is known to within 5%. A practical shortcut: round your final answer to match the least precise input. Scientific Notation and Significant Digits Normalized scientific notation expresses a number as: $$s \times 10^n$$ where $1 \le s < 10$ and $n$ is an integer. The significand $s$ contains all significant digits. In scientific notation, every digit displayed in the significand is significant. For instance, $1.25 \times 10^3$ has 3 significant digits, while $1.250 \times 10^3$ has 4 significant digits. <extrainfo> Exponentiation and Logarithms in Real Arithmetic For positive bases, exponentiation (raising a number to a power) is closed in the real numbers. Logarithms, the inverse operation, are defined only for positive arguments and bases not equal to 1. These properties ensure that certain operations always produce valid results within the real number system. Exponentiation by Squaring When computing large powers, naive repeated multiplication is inefficient. Exponentiation by squaring accelerates this process by using the fact that $x^{2n} = (x^n)^2$ and $x^{2n+1} = x \cdot x^{2n}$. This reduces the number of multiplications needed from O(n) to O(log n). </extrainfo> Floating-Point Arithmetic in Practice Non-Associativity of Addition One of the most surprising properties of floating-point arithmetic is that addition is not associative. In exact arithmetic, $(a + b) + c = a + (b + c)$ always holds. But with rounding errors: $$(a + b) + c \neq a + (b + c)$$ This occurs because rounding happens after each operation. Adding a very large number and a very small number can result in the small number being "lost" to rounding, and the order in which you combine numbers affects whether this happens. <extrainfo> IEEE 754 Standard The IEEE 754 standard (published by the Institute of Electrical and Electronics Engineers) defines how computers represent and perform floating-point arithmetic. It specifies formats (like 32-bit and 64-bit), how rounding is done, how special values (like infinity and "not a number") are handled, and how errors should be reported. This standardization ensures that floating-point arithmetic behaves consistently across different computers. Arbitrary-Precision Arithmetic When speed is not a constraint, arbitrary-precision arithmetic can be used, where precision is limited only by available computer memory rather than a fixed number of bits. This is valuable for applications requiring extremely accurate results, such as cryptography or mathematical research. Languages and libraries like Python and GMP (GNU Multiple Precision) support this. </extrainfo>