Notebook 4 Flashcards
Number
String of digits in a base b
Bits
Binary Digits
Floating Point Storage
Encode numbers with fractional parts in a systematic way using fixed-size encoding.
Systematic encoding
sign, significand, exponent
Systematic encoding
sign, significand, exponent
Rounding Error
Not being able to store the exact solution due to too much digits
Binary Encoding
IEEE 754 Standard
IEEE Double-precision
64 bits total
1 sign bit
53 bit significant
11 bit exponent
IEEE Single-precision
32 bits total
1 sign bit
24 bit significand
8 bit exponent
Analyzing Floating-point programs
f(x) = Function to compute
alg(x) = program to compute f(x)
lalg(x)- f(x)l - forward stability analysis
“small”
forward stable
Forward Stable
|alg(x) - f(x)| <= “small”
Backward Stable
alg(x) = f(x+deltax),|deltax|
Taylor’s theorem
f(x+deltax) = f(x) + deltax x first deriv x
Small backward error ->
forward error small