Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
Scribd Logo
Upload

Welcome to Scribd!

  • 338 views

    Extended Master Theorem

    Uploaded by

    Ekta Singh
    This document presents an extension to the Master Theorem to fill gaps in cases where the theorem does not apply. It introduces 4 new cases: 1) If f(n) = O(n^E(logbn)^α) with α < -1, then T(n) = Θ(n^E). 2) If f(n) = Θ(n^E(logbn)^-1), then T(n) = Θ(n^E logblogb(n)). 3) If f(n) = Θ(n^E(logbn)^α) with α > -1, then T(n) = Θ(n

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd

    Extended Master Theorem

    Uploaded by

    Ekta Singh
    338 views3 pages
    This document presents an extension to the Master Theorem to fill gaps in cases where the theorem does not apply. It introduces 4 new cases: 1) If f(n) = O(n^E(logbn)^α) with α < -1, then T(n) = Θ(n^E). 2) If f(n) = Θ(n^E(logbn)^-1), then T(n) = Θ(n^E logblogb(n)). 3) If f(n) = Θ(n^E(logbn)^α) with α > -1, then T(n) = Θ(n

    Original Description:

    Algorithms : Extended Master Theorem

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    This document presents an extension to the Master Theorem to fill gaps in cases where the theorem does not apply. It introduces 4 new cases: 1) If f(n) = O(n^E(logbn)^α) with α < -1, then T(n) = Θ(n^E). 2) If f(n) = Θ(n^E(logbn)^-1), then T(n) = Θ(n^E logblogb(n)). 3) If f(n) = Θ(n^E(logbn)^α) with α > -1, then T(n) = Θ(n

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    338 views3 pages

    Extended Master Theorem

    Uploaded by

    Ekta Singh
    This document presents an extension to the Master Theorem to fill gaps in cases where the theorem does not apply. It introduces 4 new cases: 1) If f(n) = O(n^E(logbn)^α) with α < -1, then T(n) = Θ(n^E). 2) If f(n) = Θ(n^E(logbn)^-1), then T(n) = Θ(n^E logblogb(n)). 3) If f(n) = Θ(n^E(logbn)^α) with α > -1, then T(n) = Θ(n

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    of 3

    An Extension to The Master Theorem

    In the Master Theorem, as given in the textbook and previous


    handout, there is a gap between cases (1) and (2), and a gap
    between cases (2) and (3).
    For example, if a = b = 2 and f (n) = n/lg(n) or f (n) = n lg(n), none
    of the cases apply. The extension below partially fills these gaps.
    def

    THEOREM (Extension of Master Theorem) If a, b, E = logb(a),


    and f (n) are as in the Master Theorem, the recurrence
    T(n) = aT(n /b) + f (n), T(1) = d,
    has solution as follows:
    1') If f (n) = O( n E (logbn)) with < 1, then T(n) = (n E).
    2') If f (n) = ( n E (logbn)1) , then T(n) = ( n E logblogb(n)) .
    3') If f (n) = ( n E (logbn)) with > 1, then
    T(n) = ( nE (logbn)+1) .
    4') [same as in Master Theorem] If f (n) = ( n E+ ) for some
    > 0, then T(n) = (f (n)), provided there is a constant c
    with c < 1 such that
    a f (n/b) c f (n) for all n sufficiently large.
    Note: (1') above includes case (1) of the Master Theorem.
    (3') above with = 0 is case (2) in the Master Theorem.

    We make use of the fact below, which follows from the close
    connection between sums and integrals.

    LEMMA. i=1 i converges if < 1 and diverges otherwise.


    n
    n
    i=1 i ln(n)+ if = 1, and i=1 i n +1 /(+1) if > 1.
    Proof of the extended Master Theorem when n is a power of b.
    Case (4) is exactly as in the Master Theorem, so we consider only
    (1), (2), and (3). In case 1, f (n) (nE (logbn)). In cases (2) and
    (3), f (n) = (nE (logbn)) for some .
    Let n = bk, so k = logb(n). From the previous handout, we know
    that
    T(n) = f (n) + af (n/b) + a2 f (n/b2) + ... + ak1 f (n/bk1) + ak d.
    Putting f (n) cnE (logbn) for some constant c, we get
    T(n) cnE (logbn)
    + ac(n/b)E (logb(n/b))
    + a2 c(n/b2)E (logb(n/b2))
    + ...
    + a k1c(n/bk1)E (logb(n/bk1))
    + ak d
    (In case 1, this is just an upper bound for T(n).)

    Note ak = nE. Also n = bk, so logb (n/bi ) = logb(bk i ) = k i. Finally,


    note a = b E , so in ai c(n/bi )E in the formula above, a i / biE = 1.
    With these simplifications, our formula becomes
    T(n) cnE k + cnE (k1) + cnE (k2) + ... + cnE 1 + dnE
    k

    = cnE i=1 i + dnE

    If < 1, then 1 i=1 i < c', where c' = i=1 i = some


    constant. So at worst T(n) (cc'+d) nE = (nE). But in the
    handout on the Master Theorem we remarked that T(n) can never
    be less than (nE), since the bottom level alone requires this much
    time.
    k

    If = 1, then i=1 i ln(k) = ln logb (n) = q logblogb(n) for some


    constant q, so
    T(n) cq nE logblogb(n) + dnE = (nE logblogb(n)).
    k

    If > 1, then +1 > 0, and i=1 i k +1 /(+1) =


    logb (n)+1 / (+1), so
    T(n) c nE logb (n)+1 / (+1) + dnE = (nE (logb (n))+1 ).

    You might also like