Skip to content

Little o: negligible relative growth

Textbook

For sufficiently large natural $n$, the functions satisfy $f(n)\ge0$ and $g(n)>0$. The statement $f\in o(g)$ means $\lim_{n\to\infty}f(n)/g(n)=0$. It compares relative size rather than absolute values.

Equivalently, for every $\varepsilon>0$ some threshold $n_0$ ensures $f(n)\le\varepsilon g(n)$ afterward. Unlike big O, the multiplier can be demanded arbitrarily small by moving the start of the tail.

For example, $n\in o(n^2)$ because the ratio is $1/n\to0$. The function $n$ itself does not decrease or tend to zero. But $3n\notin o(n)$: its ratio stays 3, although $3n\in O(n)$ is true.

Negligibility here is asymptotic. At small inputs, constants can make a lower-order contribution numerically significant. Dropping a term in a concrete calculation requires checking its relative contribution at the scale being used.

The threshold $n_0$ may be selected anew for each $\varepsilon$, but the condition must then hold for every larger $n$. Isolated favorable ratios are insufficient. For positive functions, multiplying $f$ by any fixed positive constant preserves membership in $o(g)$: the constant changes the required threshold, while the ratio’s zero limit remains zero.

Definitions and notation: Mathematics for Computer Science.

Related concepts: Limits of sequences, Big O: an upper growth bound.

In plain language

An expense can grow in absolute currency while taking a shrinking share of a budget. Little o requires that share to tend to zero and remain arbitrarily small throughout a sufficiently distant tail. A linear term has this fate beside a quadratic even though the linear term itself keeps growing. “Negligible” concerns relative scales: pocket money can be astronomical for a large enough owner, and one favorable month does not establish a trend.

Example

The constructed example compares $f(n)=n$ and $g(n)=n^2$ through both values and their ratio.

Little o: negligible relative growth

Input 1

The functions are equal here; the ratio is 1.

Step 1 of 4: Input 1

Read all steps
  1. Input 1

    The functions are equal here; the ratio is 1.

    f; g; ratio
    111current
  2. Input 10

    The numerator grows, but its share shrinks to 0.1.

    f; g; ratio
    101000.1current
  3. Input 100

    The linear part is now one hundredth of the quadratic.

    f; g; ratio
    100100000.01included
  4. Any positive share

    For any positive tolerance, n above its reciprocal is sufficient.

    Tolerance; sufficient threshold
    0.001currentn > 1000included

The requirement “below one percent” gives $1/n<0.01$, hence $n>100$. The row at $n=100$ is exactly one percent, so it has not yet met the strict tolerance. Replacing the numerator with $1000n$ requires $n>100000$. It still has lower order, yet can remain significant at moderate inputs. A limit statement therefore does not supply a ready numerical bound for any arbitrary input.

The criterion is the limit of the ratio. Joint growth or an ordinary big-O bound is not enough for little o.

The original text of this free article is published under CC BY-NC-ND 4.0.

Previous / next

Previous
Asymptotic equivalence
Next
Big O: an upper growth bound
  • Asymptotic equivalence

    When two functions differ by a vanishing relative fraction.

    Read article
    Algorithms
  • Big O: an upper growth bound

    How a constant factor and a threshold define an asymptotic upper bound.

    Read article
    Algorithms
  • Big Omega: a lower growth bound

    How to show a function eventually stays above a chosen growth scale.

    Read article
    Algorithms