Line Search Methods for

Unconstrained Optimisation

Lecture 8, Numerical Linear Algebra and Optimisation

Oxford University Computing Laboratory, MT 2007
Dr Raphael Hauser ([email protected])
The Generic Framework
For the purposes of this lecture we consider the unconstrained minimisation
problem
(UCM) minn f (x),
x∈R

where f ∈ C 1(Rn, R) with Lipschitz continous gradient g(x).

• In practice, these smoothness assumptions are sometimes violated, but

the algorithms we will develop are still observed to work well.

• The algorithms we will construct have the common feature that, starting
from an initial educated guess x0 ∈ Rn for a solution of (UCM), a sequence
of solutions (xk )N ⊂ Rn is produced such that
xk → x∗ ∈ Rn
such that the first and second order necessary optimality conditions
g(x∗) = 0,
H(x∗)  0 (positive semidefiniteness)
are satisfied.
 • We usually wish to make progress towards solving (UCM) in every itera-
tion, that is, we will construct xk+1 so that
f (xk+1 ) < f (xk )
(descent methods).

• In practice we cannot usually compute x∗ precisely (i.e., give a symbolic

representation of it, see the LP lecture!), but we have to stop with a xk
sufficiently close to x∗ .

• Optimality conditions are still useful, in that they serve as a stopping

criterion when they are satisfied to within a predetermined error tolerance.

• Finally, we wish to construct (xk )N such that convergence to x∗ takes

place at a rapid rate, so that few iterations are needed until the stopping
criterion is satisfied. This has to be counterbalanced with the computa-
tional cost per iteration, as there typically is a tradeoff
faster convergence ⇔ higher computational cost per iteration.

We write f k = f (xk ), g k = g(xk ), and H k = H(xk ).

Generic Line Search Method:

1. Pick an initial iterate x0 by educated guess, set k = 0.

2. Until xk has converged,

i) Calculate a search direction pk from xk , ensuring that this direction
is a descent direction, that is,
[g k ]Tpk < 0 if g k 6= 0,
so that for small enough steps away from xk in the direction pk the
objective function will be reduced.
ii) Calculate a suitable steplength αk > 0 so that
f (xk + αk pk ) < f k .
The computation of αk is called line search, and this is usually an
inner iterative loop.
iii) Set xk+1 = xk + αk pk .

Actual methods differ from one another in how steps i) and ii) are computed.
Computing a Step Length αk

The challenges in finding a good αk are both in avoiding that

the step length is too long,

2.5

1.5

0.5

−2 −1.5 −1 −0.5 0 0.5 1 1.5 2

(the objective function f (x) = x2 and the iterates xk+1 = xk + αk pk generated

by the descent directions pk = (−1)k+1 and steps αk = 2+3/2k+1 from x0 = 2)
or too short,

2.5

1.5

0.5

−2 −1.5 −1 −0.5 0 0.5 1 1.5 2

(the objective function f (x) = x2 and the iterates xk+1 = xk + αk pk generated by the descent
directions pk = −1 and steps αk = 1/2k+1 from x0 = 2).
Exact Line Search:

In early days, αk was picked to minimize

(ELS) min f (xk + αpk )
α
s.t. α ≥ 0.
Although usable, this method is not considered cost effective.

Inexact Line Search Methods:

• Formulate a criterion that assures that steps are neither too long nor too
short.

• Pick a good initial stepsize.

• Construct sequence of updates that satisfy the above criterion after very
few steps.
Backtracking Line Search:

1. Given αinit > 0 (e.g., αinit = 1), let α(0) = αinit and l = 0.

2. Until f (xk + α(l)pk )“<”f k ,

i) set α(l+1) = τ α(l), where τ ∈ (0, 1) is fixed (e.g., τ = 12 ),
ii) increment l by 1.

3. Set αk = α(l) .

This method prevents the step from getting too small, but it does not prevent
steps that are too long relative to the decrease in f .

To improve the method, we need to tighten the requirement

f (xk + α(l)pk )“<”f k .
To prevent long steps relative to the decrease in f , we require the Armijo
condition
f (xk + αk pk ) ≤ f (xk ) + αk β · [g k ]Tpk
for some fixed β ∈ (0, 1) (e.g., β = 0.1 or even β = 0.0001).

That is to say, we require that the achieved reduction if f be at least a fixed

fraction β of the reduction promised by the first-oder Taylor approximation
of f at xk .
0.14

0.12

0.1

0.08

0.06

0.04

0.02

−0.02

−0.04

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2

Backtracking-Armijo Line Search:

1. Given αinit > 0 (e.g., αinit = 1), let α(0) = αinit and l = 0.

2. Until f (xk + α(l)pk ) ≤ f (xk ) + α(l)β · [g k ]Tpk ,

1 ),
i) set α(l+1) = τ α(l), where τ ∈ (0, 1) is fixed (e.g., τ = 2

ii) increment l by 1.

3. Set αk = α(l).
Theorem 1 (Termination of Backtracking-Armijo). Let f ∈ C 1 with gradient
g(x) that is Lipschitz continuous with constant γ k at xk , and let pk be a
descent direction at xk . Then, for fixed β ∈ (0, 1),

i) the Armijo condition f (xk + αpk ) ≤ f k + αβ · [g k ]Tpk is satisfied for all

α ∈ [0, αkmax], where
2(β − 1)[g k ]Tpk
αkmax = ,
γ k kpk k22

ii) and furthermore, for fixed τ ∈ (0, 1) the stepsize generated by the backtracking-
Armijo line search terminates with
k ] T pk
 
2τ (β − 1)[g
αk ≥ min αinit, .
γ k kpk k22

We remark that in practice γ k is not known. Therefore, we cannot simply

compute αkmax and αk via the explicit formulas given by the theorem, and we
still need the algorithm on the previous slide.
Theorem 2 (Convergence of Generic LSM with B-A Steps).
Let the gradient g of f ∈ C 1 be uniformly Lipschitz continuous
on Rn. Then, for the iterates generated by the Generic Line
Search Method with Backtracking-Armijo step lengths, one of
the following situations occurs,

i) g k = 0 for some finite k,

ii) limk→∞ f k = −∞,

[g k ]Tpk
 
k T k
iii) limk→∞ min |[g ] p |, = 0.
kpk k2
Computing a Search Direction pk

Method of Steepest Descent:

The most straight-forward choice of a search direction, pk = −g k , is called
steepest-descent direction.

• pk is a descent direction.

• pk solves the problem

k T
min p ∈ Rn mL k k
k (x + p) = f + [g ] p
s.t. kpk2 = kg k k2 .

• pk is cheap to compute.

Any method that uses the steepest-descent direction as a search direction is

a method of steepest descent.
Intuitively, it would seem that pk is the best search-direction one can find. If
that were true then much of optimisation theory would not exist!
Theorem 3 (Global Convergence of Steepest Descent). Let the
gradient g of f ∈ C 1 be uniformly Lipschitz continuous on Rn.
Then, for the iterates generated by the Generic LSM with B-A
steps and steepest-descent search directions, one of the following
situations occurs,

i) g k = 0 for some finite k,

ii) limk→∞ f k = −∞,

iii) limk→∞ g k = 0.
Advantages and disadvantages of steepest descent:

⊕ Globally convergent (converges to a local minimiser from any

starting point x0).

⊕ Many other methods switch to steepest descent when they

do not make sufficient progress.

⊖ Not scale invariant (changing the inner product on Rn changes

the notion of gradient!).

⊖ Convergence is usually very (very!) slow (linear).

⊖ Numerically, it is often not convergent at all.

 1.5

0.5

−0.5
−1 −0.8 −0.6 −0.4 −0.2 0 0.2 0.4 0.6 0.8 1

Contours for the objective function f (x, y) = 10(y − x2 )2 + (x − 1)2 (Rosenbrock function),
and the iterates generated by the generic line search steepest-descent method.
More General Descent Methods:

Let B k be a symmetric, positive definite matrix, and define the

search direction pk as the solution to the linear system

B k pk = −g k

• pk is a descent direction, since

[g k ]Tpk = −[g k ]T[B k ]−1g k < 0.

• pk solves the problem

Q 1
min mk (xk + p) = f k + [g k ]Tp + pTB k p.
p∈Rn 2
• pk corresponds to the steepest descent direction if the norm
q
kxkB k := xT B k x
is used on Rn instead of the canonical Euclidean norm. This
change of metric can be seen as preconditioning that can be
chosen so as to speed up the steepest descent method.

• If the Hessian H k of f at xk is positive definite, and B k = H k ,

this is Newton’s method.

• If B k changes at every iterate xk , a method based on the

search direction pk is called variable metric method. In par-
ticular, Newton’s method is a variable metric method.
Theorem 4 (Global Convergence of More General Descent Di-
rection Methods). Let the gradient g of f ∈ C 1 be uniformly
Lipschitz continuous on Rn . Then, for the iterates generated by
the Generic LSM with B-A steps and search directions defined
by B k pk = −g k , one of the following situations occurs,

i) g k = 0 for some finite k,

ii) limk→∞ f k = −∞,

iii) limk→∞ g k = 0,

provided that the eigenvalues of B k are uniformly bounded above,

and uniformly bounded away from zero.
Theorem 5 (Local Convergence of Newton’s Method). Let the Hessian H of
f ∈ C 2 be uniformly Lipschitz continuous on Rn. Let iterates xk be generated
via the Generic LSM with B-A steps using αinit = 1 and β < 12 , and using
the Newton search direction nk , defined by H k nk = −g k . If (xk )N has an
accumulation point x∗ where H(x∗) ≻ 0 (positive definite) then

i) αk = 1 for all k large enough,

ii) limk→∞ xk = x∗ ,

iii) the sequence converges Q-quadratically, that is, there exists κ > 0 such
that
kxk+1 − x∗k
lim ≤ κ.
k→∞ kxk − x∗ k2

The mechanism that makes Theorem 5 work is that once the sequence
(xk )N enters a certain domain of attraction of x∗, it cannot escape again
and quadratic convergence to x∗ commences.
Note that this is only a local convergence result, that is, Newton’s method is
not guaranteed to converge to a local minimiser from all starting points.
The fast convergence of Newton’s method becomes apparent
when we apply it to the Rosenbrock function:
1.5

0.5

−0.5
−1 −0.8 −0.6 −0.4 −0.2 0 0.2 0.4 0.6 0.8 1

Contours for the objective function f (x, y) = 10(y −x2 )2 +(x−1)2 , and the iterates generated
by the Generic Linesearch Newton method.
Modified Newton Methods:

The use of B k = H k makes only sense at iterates xk where

H k ≻ 0. Since this is usually not guaranteed to always be the
case, we modify the method as follows,

• Choose M k  0 so that H k + M k is “sufficiently” positive

definite, with M k = 0 if H k itself is sufficiently positive defi-
nite.

• Set B k = H k + M k and solve B k pk = −g k .

The regularisation term M k is typically chosen as one of the following,

• If H k has the spectral decomposition H k = Qk Λk [Qk ]T, then

H k + M k = Qk max(ε I, |Dk |)[Qk ]T.

• M k = max(0, −λmin (H k )) I.

• Modified Cholesky method:

1. Compute a factorisation P H k P T = LBLT, where P is a permutation
matrix, L a unit lower triangular matrix, and B a block diagonal matrix
with blocks of size 1 or 2.
2. Choose a matrix F such that B + F is sufficiently positive definite.
3. Let H k + M k = P T L(B + F )LTP .
Other Modifications of Newton’s Method:

1. Build a cheap approximation B k to H k :

• Quasi-Newton approximation (BFGS, SR1, etc.),

• or use finite-difference approximation.

2. Instead of solving B k pk = −g k for pk , if B k ≻ 0 approximately

solve the convex quadratic programming problem
1
(QP) pk ≈ arg min f k + pTg k + pTBp.
p∈Rn 2
The conjugate gradient method is a good solver for step 2:

1. Set p(0) = 0, g (0) = g k , d(0) = −g k , and i = 0.

2. Until g (i) is sufficiently small or i = n, repeat

kg (i) k22
i) α(i) = [d(i) ]T B k d(i)
,

ii) p(i+1) = p(i) + α(i) d(i) ,

iii) g (i+1) = g (i) + α(i) B k d(i) ,
kg (i+1) k22
iv) β (i) = kg (i) k22
,

v) d(i+1) = −g (i+1) + β (i) d(i) ,

vi) increment i by 1.

3. Output pk ≈ p(i) .
Important features of the conjugate gradient method:

• [g k ]Tp(i) < 0 for all i, that is, the algorithm always stops with
a descent direction as an approximation to pk .

• Each iteration is cheap, as it only requires the computation

of matrix-vector and vector-vector products.

• Usually, p(i) is a good approximation of pk well before i = n.