Solutions of Assignment of Week 07
2 −1
Problem 1. Let f : R → R be a function given by f (x1 , x2 ) = 2x21
− 2x1 − x2 − x22 . 1
Consider d = ,
2
−1 2
d2 = , and d3 = in R2 . Then, which of the following is (are) descent direction(s) of f at (0, 0)?
−2 1
(a) only d3
(b) all three d1 , d2 , and d3
(c) only d1 and d3
(d) only d2 and d3
Problem 2. Choose the incorrect statement from below.
(a) Let f : S ⊆ Rn → R be a function on the nonempty convex set S and f be differentiable on S. If
∇f (x̄) = 0 for some x̄ ∈ S, then x̄ is a global minimizer of f on S.
(b) Let f : Rn → R be twice differentiable at x̄ ∈ Rn . Suppose that x̄ is a local minimizer of f . Then,
∇f (x̄) = 0 and ∇2 f (x̄) is positive semidefinite.
(c) Let Rn → R be twice differentiable at x̄ ∈ Rn . If ∇f (x̄) = 0 and ∇2 f (x̄) is positive definite, then x̄ is a
strict local minimizer of f .
(d) If x̄ ∈ Rn is a local minimizer of the convex function f : Rn → R, then x̄ is a global minimizer of f .
Problem 3. Let f : R2 → R be a function given by f (x1 , x2 ) = x21 − x22 − 2x1 . Then,
(a) (0, 0) is a stationary point of f
(b) (1, 0) is a point of local maximum of f
(c) (1, 0) is a saddle point of f
(d) (0, 1) is a stationary point of f
Problem 4. Which of the following statements is the first order necessary condition?
(a) Let f : S → R be a differentiable function on a nonempty open set S ⊆ Rn . If x∗ ∈ S is a local minimizer
of f , then ∇f (x∗ ) = 0.
(b) Let f : S → R be a differentiable function on a nonempty open set S ⊆ Rn . If x∗ ∈ S is a local minimizer
of f , then all components of ∇f (x∗ ) are negative.
(c) Let f : S → R be a differentiable function on a nonempty open set S ⊆ Rn . If x∗ ∈ S is a local minimizer
of f , then all components of ∇f (x∗ ) are positive.
(d) Let f : S → R be a twice differentiable function on a nonempty open set S ⊆ Rn . If x∗ ∈ S is a local
minimizer of f , then all components of ∇2 f (x∗ ) are zero.
Problem 5. Consider the two statements below.
S1: Let f : S ⊆ Rn → R be a function and x̄ ∈ S̃ ⊆ S be a global minimizer of f over S. Then, x̄ is a global
minimizer of f over S̃.
S2: Let f : S ⊆ Rn → R be a function and x̄ ∈ S̃ ⊆ S be a local minimizer of f over S̃. Then, x̄ is a local
minimizer of f over S.
Then,
1
(a) only S1 is true
(b) only S2 is true
(c) both S1 and S2 are true
(d) both S1 and S2 are not true
Problem 6. Choose the incorrect statement from below.
(a) General descent algorithm can converge to a saddle point.
(b) Descent method can converge to a nonstrict local maximum.
(c) General descent algorithm can generate a convergent sequence, which converges to a local strict maximum
for any initial point.
(d) Global convergence of general descent algorithm can be proved using both inexact and exact line search
rules.
Problem 7. An optimization algorithm is said to have descent property if for any sequence {xk } generated
by the algorithm, we have
(a) f (xk+1 ) = 0
(b) f (xk+1 ) ≥ f (xk ) and ∇f (xk ) ̸= 0
(c) f (xk+1 ) < f (xk ) and ∇f (xk ) ̸= 0
(d) f (xk+1 ) = f (xk )
Problem 8. Which of the following statements is second order sufficient condition?
(a) Let f : S ⊆ Rn → R be a convex function on the nonempty convex set S and differentiable on S. If
∇f (x̄) = 0 for some x ∈ S, then x̄ is a global minimizer of f on S.
(b) Let f : Rn → R be twice differentiable function at x̄ ∈ Rn . Suppose that x̄ is a local minimizer of f .
Then, ∇f (x̄) = 0 and ∇2 f (x̄) is positive definite.
(c) Let f : Rn → R be twice differentiable function at x̄ ∈ Rn . If ∇f (x̄) = 0 and ∇2 f (x̄) is positive definite,
then x̄ is a strict local minimizer of f .
(d) Let f : Rn → R be twice differentiable function at x̄ ∈ Rn . Suppose that x̄ is a local minimizer of f .
Then, ∇f (x̄) = 0 and ∇2 f (x̄) is positive semidefinite.
Problem 9. Let f and g be two functions given by f (x1 , x2 ) = x21 +x32 and g(x1 , x2 ) = x21 +x42 , respectively,
where (x1 , x2 ) ∈ R2 . Then,
(a) (0, 0) is a local minimizer of both f and g
(b) (0, 0) is a local minimizer of f only
(c) (0, 0) is a local minimizer of g only
(d) (0, 0) is not a local minimizer of both f and g
Problem 10. Let f : Rn → R be a continuously differentiable function. Then, Wolfe-Powell line search
conditions, to find αk , at xk with descent direction dk are
(a) f (xk + αk dk ) ≤ f (xk ) + αk c1 ∇f (xk )⊤ dk and ∇f (xk + αk dk )⊤ dk ≥ c2 ∇f (xk )⊤ dk , where 0 < c1 < c2 < 1
(b) f (xk + αk dk ) ≤ f (xk ) + αk c1 ∇f (xk )⊤ dk and ∇f (xk + αk dk )⊤ dk ≤ c2 ∇f (xk )⊤ dk , where 0 < c1 < c2 < 1
(c) f (xk + αk dk ) ≥ f (xk ) + αk c1 ∇f (xk )⊤ dk and ∇f (xk + αk dk )⊤ dk ≤ c2 ∇f (xk )⊤ dk , where 0 < c1 < c2 < 1
(d) f (xk + αk dk ) ≥ f (xk ) + αk c1 ∇f (xk )⊤ dk and ∇f (xk + αk dk )⊤ dk ≥ c2 ∇f (xk )⊤ dk , where 0 < c1 < c2 < 1