Substitution Method and Recursion Tree Method for:
T(n) = 2/7 If n=1
T(n) = 2T(n/3)+n^2 Otherwise
How to solve such a system of equations?
The equation clearly only can be solved when n is a power of 3. Substituting n by a new variable k^3 gives a simpler equation that can be solved by sympy, python's symbolic mathematics library:
from sympy import Function, rsolve, S, symbols, Eq
k, n = symbols("k n", integer=True, positive=True)
f = Function('f')
g = Function('g')
# T(n) = 2T(n/3)+n^2 T(1) = 2/7
T = Eq(f(n), 2 * f(n / 3) + n * n)
Tk = T.subs({n: 3 ** k, f(n): g(k), f(n / 3): g(k - 1)})
s = rsolve(Tk, g(k), {g(0): 2 / S(7)})
print("solution for k:", s.cancel())
for k in range(0, 11):
print(f"k={k}, n={3 ** k}, T(n)={-2 ** k + 3 ** (2 * k + 2) / S(7)}")
This results in following formula for k: -2k + 32*k+2/7
The values for k=0..10 are:
k=0, n=1, T(n)=2/7
k=1, n=3, T(n)=67/7
k=2, n=9, T(n)=701/7
k=3, n=27, T(n)=6505/7
k=4, n=81, T(n)=58937/7
k=5, n=243, T(n)=531217/7
k=6, n=729, T(n)=4782521/7
k=7, n=2187, T(n)=43045825/7
k=8, n=6561, T(n)=387418697/7
k=9, n=19683, T(n)=3486780817/7
k=10, n=59049, T(n)=31381052441/7
Related
I am trying to minimize a nonlinear function with nonlinear inequality constraints with NLopt and JuMP.
In my test code below, I am minimizing a function with a known global minima.
Local optimizers such as LD_MMA fails to find this global minima, so I am trying to use global optimizers of NLopt that allow nonlinear inequality constraintes.
However, when I check my termination status, it says “termination_status(model) = MathOptInterface.OTHER_ERROR”. I am not sure which part of my code to check for this error.
What could be the cause?
I am using JuMP since in the future I plan to use other solvers such as KNITRO as well, but should I rather use the NLopt syntax?
Below is my code:
# THIS IS A CODE TO SOLVE FOR THE TOYMODEL
# THE EQUILIBRIUM IS CHARACTERIZED BY A NONLINEAR SYSTEM OF ODEs OF INCREASING FUCTIONS B(x) and S(y)
# THE GOAL IS TO APPROXIMATE B(x) and S(y) WITH POLYNOMIALS
# FIND THE POLYNOMIAL COEFFICIENTS THAT MINIMIZE THE LEAST SQUARES OF THE EQUILIBRIUM EQUATIONS
# load packages
using Roots, NLopt, JuMP
# model primitives and other parameters
k = .5 # equal split
d = 1 # degree of polynomial
nparam = 2*d+2 # number of parameters to estimate
m = 10 # number of grids
m -= 1
vGrid = range(0,1,m) # discretize values
c1 = 0 # lower bound for B'() and S'()
c2 = 2 # lower and upper bounds for offers
c3 = 1 # lower and upper bounds for the parameters to be estimated
# objective function to be minimized
function obj(α::T...) where {T<:Real}
# split parameters
αb = α[1:d+1] # coefficients for B(x)
αs = α[d+2:end] # coefficients for S(y)
# define B(x), B'(x), S(y), and S'(y)
B(v) = sum([αb[i] * v .^ (i-1) for i in 1:d+1])
B1(v) = sum([αb[i] * (i-1) * v ^ (i-2) for i in 2:d+1])
S(v) = sum([αs[i] * v .^ (i-1) for i in 1:d+1])
S1(v) = sum([αs[i] * (i-1) * v ^ (i-2) for i in 2:d+1])
# the equilibrium is characterized by the following first order conditions
#FOCb(y) = B(k * y * S1(y) + S(y)) - S(y)
#FOCs(x) = S(- (1-k) * (1-x) * B1(x) + B(x)) - B(x)
function FOCb(y)
sy = S(y)
binv = find_zero(q -> B(q) - sy, (-c2, c2))
return k * y * S1(y) + sy - binv
end
function FOCs(x)
bx = B(x)
sinv = find_zero(q -> S(q) - bx, (-c2, c2))
return (1-k) * (1-x) * B1(x) - B(x) + sinv
end
# evaluate the FOCs at each grid point and return the sum of squares
Eb = [FOCb(y) for y in vGrid]
Es = [FOCs(x) for x in vGrid]
E = [Eb; Es]
return E' * E
end
# this is the actual global minimum
αa = [1/12, 2/3, 1/4, 2/3]
obj(αa...)
# do optimization
model = Model(NLopt.Optimizer)
set_optimizer_attribute(model, "algorithm", :GN_ISRES)
#variable(model, -c3 <= α[1:nparam] <= c3)
#NLconstraint(model, [j = 1:m], sum(α[i] * (i-1) * vGrid[j] ^ (i-2) for i in 2:d+1) >= c1) # B should be increasing
#NLconstraint(model, [j = 1:m], sum(α[d+1+i] * (i-1) * vGrid[j] ^ (i-2) for i in 2:d+1) >= c1) # S should be increasing
register(model, :obj, nparam, obj, autodiff=true)
#NLobjective(model, Min, obj(α...))
println("")
println("Initial values:")
for i in 1:nparam
set_start_value(α[i], αa[i]+rand()*.1)
println(start_value(α[i]))
end
JuMP.optimize!(model)
println("")
#show termination_status(model)
#show objective_value(model)
println("")
println("Solution:")
sol = [value(α[i]) for i in 1:nparam]
My output:
Initial values:
0.11233072522513032
0.7631843020124309
0.3331559403539963
0.7161240026812674
termination_status(model) = MathOptInterface.OTHER_ERROR
objective_value(model) = 0.19116585196576466
Solution:
4-element Vector{Float64}:
0.11233072522513032
0.7631843020124309
0.3331559403539963
0.7161240026812674
I answered on the Julia forum: https://discourse.julialang.org/t/mathoptinterface-other-error-when-trying-to-use-isres-of-nlopt-through-jump/87420/2.
Posting my answer for posterity:
You have multiple issues:
range(0,1,m) should be range(0,1; length = m) (how did this work otherwise?) This is true for Julia 1.6. The range(start, stop, length) method was added for Julia v1.8
Sometimes your objective function errors because the root doesn't exist. If I run with Ipopt, I get
ERROR: ArgumentError: The interval [a,b] is not a bracketing interval.
You need f(a) and f(b) to have different signs (f(a) * f(b) < 0).
Consider a different bracket or try fzero(f, c) with an initial guess c.
Here's what I would do:
using JuMP
import Ipopt
import Roots
function main()
k, d, c1, c2, c3, m = 0.5, 1, 0, 2, 1, 10
nparam = 2 * d + 2
m -= 1
vGrid = range(0, 1; length = m)
function obj(α::T...) where {T<:Real}
αb, αs = α[1:d+1], α[d+2:end]
B(v) = sum(αb[i] * v^(i-1) for i in 1:d+1)
B1(v) = sum(αb[i] * (i-1) * v^(i-2) for i in 2:d+1)
S(v) = sum(αs[i] * v^(i-1) for i in 1:d+1)
S1(v) = sum(αs[i] * (i-1) * v^(i-2) for i in 2:d+1)
function FOCb(y)
sy = S(y)
binv = Roots.fzero(q -> B(q) - sy, zero(T))
return k * y * S1(y) + sy - binv
end
function FOCs(x)
bx = B(x)
sinv = Roots.fzero(q -> S(q) - bx, zero(T))
return (1-k) * (1-x) * B1(x) - B(x) + sinv
end
return sum(FOCb(x)^2 + FOCs(x)^2 for x in vGrid)
end
αa = [1/12, 2/3, 1/4, 2/3]
model = Model(Ipopt.Optimizer)
#variable(model, -c3 <= α[i=1:nparam] <= c3, start = αa[i]+ 0.1 * rand())
#constraints(model, begin
[j = 1:m], sum(α[i] * (i-1) * vGrid[j]^(i-2) for i in 2:d+1) >= c1
[j = 1:m], sum(α[d+1+i] * (i-1) * vGrid[j]^(i-2) for i in 2:d+1) >= c1
end)
register(model, :obj, nparam, obj; autodiff = true)
#NLobjective(model, Min, obj(α...))
optimize!(model)
print(solution_summary(model))
return value.(α)
end
main()
Recurrence relations can be directly derived from a recursive algorithm, but
they are in a form that does not allow us to quickly determine how efficient
the algorithm is.
Please how can I solve this
T(n) = 6T(n/6) + 2n + 3 for n a power of 6 T(1) = 1 solution ?
This recurrence could be solved with rsolve from SymPy, Python's symbolic math library.
from sympy import Function, rsolve
from sympy.abc import k, n
f = Function('f')
g = Function('g')
# T(n) = 6T(n/6) + 2n + 3 for n a power of 6 T(1) = 1
T = f(n) - 6*f(n/6) - 2*n - 3
Tk = T.subs({n: 6**k, f(n): g(k), f(n/6):g(k-1)})
s = rsolve(Tk, g(k), {g(0): 1})
print ("solution for k:", s.cancel())
for k in range(0,11):
print(f"k={k}, n={6**k}, T(n)={2*6**k*k + (8*6**k - 3)//5}")
This gives:
Tk(k) = 2*6**k*k + 8*6**k/5 - 3/5 or Tk(k) = ((10k+8)6k - 3)/5
T(n) = 2*n*log(n)/log(6) + 8*n/5 - 3/5 or T(n) = ((n(10log6(n)+8) - 3)/5
First 11 values:
k=0, n=1, T(n)=1
k=1, n=6, T(n)=21
k=2, n=36, T(n)=201
k=3, n=216, T(n)=1641
k=4, n=1296, T(n)=12441
k=5, n=7776, T(n)=90201
k=6, n=46656, T(n)=634521
k=7, n=279936, T(n)=4367001
k=8, n=1679616, T(n)=29561241
k=9, n=10077696, T(n)=197522841
k=10, n=60466176, T(n)=1306069401
We can check the formulas via the recursive formulation:
def recursive_t(n):
if n == 1:
res = 1
else:
t_ndiv6 = recursive_t(n//6)
res = 6 * t_ndiv6 + 2 * n + 3
print(f"T({n})={res}")
return res
recursive_t(6**10)
This prints out the same values for the same n.
link of question
http://codeforces.com/contest/615/problem/D
link of solution is
http://codeforces.com/contest/615/submission/15260890
In below code i am not able to understand why 1 is subtracted from mod
where mod=1000000007
ll d = 1;
ll ans = 1;
for (auto x : cnt) {
ll cnt = x.se;
ll p = x.fi;
ll fp = binPow(p, (cnt + 1) * cnt / 2, MOD);
ans = binPow(ans, (cnt + 1), MOD) * binPow(fp, d, MOD) % MOD;
d = d * (x.se + 1) % (MOD - 1);//why ??
}
Apart from the fact that there is the code does not make much sense as out of context as it is, there is the little theorem of Fermat:
Whenever MOD is a prime number, as 10^9+7 is, one can reduce exponents by multiples of (MOD-1) as for any a not a multiple of MOD
a ^ (MOD-1) == 1 mod MOD.
Which means that
a^b == a ^ (b mod (MOD-1)) mod MOD.
As to the code, which is efficient for its task, consider n=m*p^e where m is composed of primes smaller than p.
Then for each factor f of m there are factors 1*f, p*f, p^2*f,...,p^e*f of n. The product over all factors of n thus is the product over
p^(0+1+2+...+e) * f^(e+1) = p^( e*(e+1)/2 ) * f^(e+1)
over all factors f of m. Putting the numbers of factors as d and the product of factors of m as ans results in the combined formula
ans = ans^( e+1 ) * p^( d*e*(e+1)/2 )
d = d*(e+1)
which can now be recursively applied to the list of prime factors and their multiplicities.
I have been asked to analyze the asymptotic time complexity of the following recursion function:
for-all k ≥ 1:
T(n) = n + T(n/2) + T(n/4) + T(n/8) + .... + T(n/2^k)
I was able to prove that:
T(n) = O(n⋅log n) and T(n) = Ω(n),
but I am looking for a tighter bound (Big Theta).
First of all:
I understand "for-all k >= 1" this way: for k = 1 to k = m where 2m-1 ≤ n ≤ 2m.
So basicly m = log₂(n) holds.
Have a look at my calculation:
T(n) = n + Σk=1,...,m T(n/2k)
= n + T(n/2) + Σk=2,...,m T(n/2k)
= n + n/2 + 2⋅Σk=2,...,m T(n/2k)
= ...
= n + Σk=1,...,m k⋅n/2k
= n + n⋅Σk=1,...,m k/2k
= n + n⋅(2 - 2-mm - 21-m)
≤ n + 2⋅n
= 3n
So T(n) is in Θ(n).
Notice:
You can also approximate Σk=1,...,m k/2k by the integral s(m) = ∫1m k/2k dk.
And here limm → ∞s(m) = 2 also holds.
I am having trouble understanding the concept of recurrences. Given you have T(n) = 2T(n/2) +1 how do you calculate the complexity of this relationship? I know in mergesort, the relationship is T(n) = 2T(n/2) + cn and you can see that you have a tree with depth log2^n and cn work at each level. But I am unsure how to proceed given a generic function. Any tutorials available that can clearly explain this?
The solution to your recurrence is T(n) ∈ Θ(n).
Let's expand the formula:
T(n) = 2*T(n/2) + 1. (Given)
T(n/2) = 2*T(n/4) + 1. (Replace n with n/2)
T(n/4) = 2*T(n/8) + 1. (Replace n with n/4)
T(n) = 2*(2*T(n/4) + 1) + 1 = 4*T(n/4) + 2 + 1. (Substitute)
T(n) = 2*(2*(2*T(n/8) + 1) + 1) + 1 = 8*T(n/8) + 4 + 2 + 1. (Substitute)
And do some observations and analysis:
We can see a pattern emerge: T(n) = 2k * T(n/2k) + (2k − 1).
Now, let k = log2 n. Then n = 2k.
Substituting, we get: T(n) = n * T(n/n) + (n − 1) = n * T(1) + n − 1.
For at least one n, we need to give T(n) a concrete value. So we suppose T(1) = 1.
Therefore, T(n) = n * 1 + n − 1 = 2*n − 1, which is in Θ(n).
Resources:
https://www.cs.duke.edu/courses/spring05/cps100/notes/slides07-4up.pdf
http://www.cs.duke.edu/~ola/ap/recurrence.html
However, for routine work, the normal way to solve these recurrences is to use the Master theorem.