Precision and accuracy

Precision: floating-point numbers always have $d$ binary digits, resolution bounded by machine epsilon

Absolute accuracy: $| \text{approx}(x) - x |$

Relative accuracy: $\frac{| \text{approx}(x) - x |}{|x|}$

Number of accurate digits: $-\log_{10} \left\vert \frac{\text{approx}(x) - x}{x} \right\vert$