PROPERTIES OF NELDER–MEAD 119
During the kth Nelder–Mead iteration, (2.12) shows that each trial point (reflection,
expansion, contraction) may be written as
z
(k)
(τ)=∆
k
t(τ), where t(τ)=
1+τ
n
, ...,
1+τ
n
, −τ
T
.(2.14)
Following the kth Nelder–Mead iteration, the (unordered) vertices of the next
simplex are the columns of ∆
k
S
k
, where S
k
is an (n +1)× (n + 1) matrix given by
I
n
(1 + τ
k
)
n
e
0
T
−τ
k
for a step of type τ and by
1(1− σ)e
T
0 σI
n
for a shrink step, with 0 being an n-dimensional zero column and I
n
being the n-
dimensional identity matrix. After being ordered at the start of iteration k + 1, the
vertices of ∆
k+1
satisfy
∆
k+1
=∆
k
T
k
, with T
k
= S
k
P
k
,(2.15)
where P
k
is a permutation matrix chosen to enforce the ordering and tie-breaking
rules (so that P
k
depends on the function values at the vertices).
The updated simplex ∆
k+1
has a disjoint interior from ∆
k
for a reflection, an
expansion, or an outside contraction, while ∆
k+1
⊆ ∆
k
for an inside contraction or a
shrink.
By the shape of a nondegenerate simplex, we mean its equivalence class under
similarity, i.e., ∆ and λ∆ have the same shape when λ>0. The shape of a simplex
is determined by its angles, or equivalently by the singular values of the associated
matrix M (2.10) after scaling so that ∆ has unit volume. The Nelder–Mead method
was deliberately designed with the idea that the simplex shapes would “adapt to the
features of the local landscape” [6]. The Nelder–Mead moves apparently permit any
simplex shape to be approximated—in particular, arbitrarily flat or needle-shaped
simplices (as in the McKinnon examples [5]) are possible.
3. Properties of the Nelder–Mead algorithm. This section establishes var-
ious basic properties of the Nelder–Mead method. Although there is a substantial
level of folklore about the Nelder–Mead method, almost no proofs have appeared in
print, so we include details here.
3.1. General results. The following properties follow immediately from the
definition of Algorithm NM.
1. A Nelder–Mead iteration requires one function evaluation when the iteration
terminates in step 2, two function evaluations when termination occurs in step 3 or
step 4, and n + 2 function evaluations if a shrink step occurs.
2. The “reflect” step is so named because the reflection point x
r
(2.4) is a
(scaled) reflection of the worst point x
n+1
around the point
¯
x on the line through
x
n+1
and
¯
x. It is a genuine reflection on this line when ρ = 1, which is the standard
choice for the reflection coefficient.
3. For general functions, a shrink step can conceivably lead to an increase in
every vertex function value except f
1
, i.e., it is possible that f
(k+1)
i
>f
(k)
i
for 2 ≤ i ≤
n + 1. In addition, observe that with an outside contraction (case 4a), the algorithm
takes a shrink step if f(x
c
) >f(x
r
), even though a new point x
r
has already been
found that strictly improves over the worst vertex, since f(x
r
) <f(x
n+1
).
Downloaded 07/06/14 to 222.195.132.86. Redistribution subject to SIAM license or copyright; see http://www.siam.org/journals/ojsa.php