Algorithms for Ellipsoids

Stephen B. Pope
Sibley School of Mechanical & Aerospace Engineering
Cornell University
Ithaca, New York 14853

Report: FDA-08-01
February 2008
 Abstract

We describe a number of algorithms to perform basic geometric operations

on ellipsoids in n spatial dimensions, for n ≥ 1. These algorithms are im-
plemented in ELL LIB, a library of Fortran subroutines. With E, E1 and
E2 being given ellipsoids, and p a given point, the tasks considered include:
determine the point in E which is closest to p or furthest from p; grow or
shrink E so that its boundary intersects p; project E onto a given affine
space; determine a separating hyperplane between E1 and E2 ; determine an
ellipsoid (of small volume) which covers E1 and E2 .
Contents
1 Introduction 3

2 Representation of ellipsoids 3

3 Summary of routines 5

4 Useful preliminary results 5

4.1 Linear transformation . . . . . . . . . . . . . . . . . . . . . . . 6
4.2 Quadratic minimization . . . . . . . . . . . . . . . . . . . . . 8
4.3 Householder matrix . . . . . . . . . . . . . . . . . . . . . . . . 8
4.4 Rank-one modification . . . . . . . . . . . . . . . . . . . . . . 9

5 Smallest and largest principal semi-axes of E 10

6 Is the point x covered by E? 12

7 Relative distance to the boundary of E 12

8 Nearest point in E to a given point 13

9 Furthest point in E to a given point 14

10 Minimum-volume ellipsoid covering E and p 14

10.1 Householder matrix algorithm . . . . . . . . . . . . . . . . . . 14
10.2 Rank-one modification algorithm . . . . . . . . . . . . . . . . 17
10.3 Behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

11 Shrink E based on a given point 18

11.1 Maximum-volume algorithm . . . . . . . . . . . . . . . . . . . 19
11.1.1 Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . 19
11.1.2 Behavior . . . . . . . . . . . . . . . . . . . . . . . . . . 19
11.2 Near-content algorithm . . . . . . . . . . . . . . . . . . . . . . 20
11.2.1 Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . 21
11.2.2 Behavior . . . . . . . . . . . . . . . . . . . . . . . . . . 22
11.3 Conservative algorithm . . . . . . . . . . . . . . . . . . . . . . 22
11.3.1 Algorithm: reduction to 2D . . . . . . . . . . . . . . . 24
11.3.2 Algorithm: solution in 2D . . . . . . . . . . . . . . . . 27

1
12 Orthogonal projection of E onto a given line 29

13 Orthogonal projection of E onto an affine space 30

14 Generate an ellipsoid which does not cover any specified

points 32

15 Separating hyperplane of two ellipsoids 34

16 Pair covering query 36

17 Shrink ellipsoid so that it is covered by a concentric ellipsoid 36

18 Ellipsoid that covers two given ellipsoids 38

18.1 Spheroid algorithm . . . . . . . . . . . . . . . . . . . . . . . . 38
18.2 Covariance algorithm . . . . . . . . . . . . . . . . . . . . . . . 39
18.3 Iterative algorithm . . . . . . . . . . . . . . . . . . . . . . . . 41
18.3.1 Stage 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
18.3.2 Stage 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
18.3.3 Stage 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
18.3.4 Stage 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
18.3.5 Stage 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
18.3.6 Stage 6 . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
18.3.7 Mutual covering . . . . . . . . . . . . . . . . . . . . . . 45
18.3.8 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . 46

19 Conclusions 46

20 Acknowledgments 46

2
1 Introduction
In this paper we describe a number of algorithms to perform basic geometric
operations on ellipsoids. These algorithms have been implemented in Fortran
routines which are contained in the library ELL LIB, which is available at
http://eccentric.mae.cornell.edu/∼tcg/ELL LIB.
Ellipsoids arise in numerous computational problems. The algorithms
and routines described here have been developed specifically for use in the
in situ adaptive tabulation (ISAT) algorithm (Pope 1997).
In Sec. 2 we consider different mathematical representations of an ellipsoid
E in <n , for n ≥ 1. For n = 1, E is a line segment; for n = 2, E is an ellipse;
for n = 3, E is an ellipsoid; and for n > 3, E is a hyper-ellipsoid. For
simplicity we generally refer to E (for all n ≥ 1) as an ellipsoid.
A summary of the routines in ELL LIB is provided in Sec. 3. Follow-
ing some preliminary results in Sec. 4, the algorithms used are described in
Secs. 5–18.

2 Representation of ellipsoids
There are many ways to represent ellipsoids, with the different ways arising
naturally in different circumstances. In this section we show the relations
between the different representations.
Let the ellipsoid E be centered at c; let the columns of the n×n orthogonal
matrix U be unit vectors in the directions of E’s principal axes; and let Σ
be the diagonal matrix (with diagonal elements Σii = σi ) such that 1/σi is
the length of the ith principal semi-axis. We assume that the principle axes
are finite and strictly positive, i.e., 0 < σi < ∞. Then E is given by
E ≡ {x | (x − c)T UΣ2 UT (x − c) ≤ 1}. (1)
This may alternatively be expressed as
E ≡ {x | kΣUT (x − c)k ≤ 1}. (2)
or, from the definition w ≡ ΣUT (x − c),
E ≡ {x | x = c + UΣ−1 w, kwk ≤ 1}. (3)
The above three definitions can be re-expressed in terms of different re-
lated matrices. Let A be the matrix

3
 A ≡ UΣ2 UT , (4)

appearing in Eq.(1). Evidently A is symmetric positive definite; its eigen-

vectors are the columns of U and its eigenvalues are λi = σi2 . We denote
by Λ = Σ2 the diagonal matrix of eigenvalues. Thus Eqs. (1)–(3) can be
1
trivially re-written by substituting Λ 2 for Σ; or less trivially in terms of A
as
E ≡ {x | (x − c)T A(x − c) ≤ 1}, (5)

1
E ≡ {x | kA 2 (x − c)k ≤ 1}, (6)

1
E ≡ {x | x = c + A− 2 w, kwk ≤ 1}. (7)
Let B be a non-singular square matrix, which we use to form A as

A = BBT , (8)

and let the SVD of B be

B = UΣVT . (9)
Note that we have from Eq.(9),

A = BBT = UΣ2 UT , (10)

consistent with Eq.(4), and showing that there is a family of matrices B

yielding the same matrix A, namely B = UΣVT for given U and Σ but
arbitrary orthogonal V. In terms of B, Eqs. (1)–(3) can be reexpressed as

E ≡ {x | (x − c)T BBT (x − c) ≤ 1}, (11)

E ≡ {x | kBT (x − c)k ≤ 1}, (12)

E ≡ {x | x = c + B−T w, kwk ≤ 1}, (13)

with a different definition of w.
The matrix B can also be factored as

B = LQ, (14)

4
where L is lower triangular with positive diagonal elements and Q is orthog-
onal. Thus we obtain
A = BBT = LLT , (15)
showing that L is the Cholesky factorization of A. The definitions of E in
terms of B apply equally in terms of L, i.e.,
E ≡ {x | (x − c)T LLT (x − c) ≤ 1}, (16)

E ≡ {x | kLT (x − c)k ≤ 1}, (17)

E ≡ {x | x = c + L−T w, kwk ≤ 1}. (18)

Computationally it is most efficient to use the Cholesky representation
of the ellipsoid in terms of c and L, and to store L in packed format. In
ELL LIB, routines with names starting ell use this representation, while those
with names starting ellu represent L in unpacked format.
In the algorithms described below, it often happens that an ellipsoid E1
given by c1 and L1 is modified to yield an ellipsoid E2 which is known in
terms of c2 and B2 (i.e., a full matrix). The representation in terms of L2
is efficiently and accurately computed by the LQ algorithm. It is stressed
that it is much more accurate to obtain L2 from the LQ algorithm than from
the Cholesky decomposition of A = B2 BT2 . All of the algorithms described
below use the LQ algorithms and avoid forming A.
The following table shows various routines in ELL LIB that can be used
to transform between one representation of an ellipsoid and another. In this
table, Ā denotes the lower triangle of the symmetric matrix A, while all
other symbols have the meanings given above.

3 Summary of routines
Table 2 summarizes the tasks performed by the principal routines in ELL LIB.
Here E, E1 and E2 denote given ellipsoids, and p denotes a given point.

4 Useful preliminary results

In this Section we give some general results that are used in the subsequent
sections.

5
 Table 1: Routines to transform between different representations
Routine From To
ell bbt2chol B L
ellu bbt2chol B L
ell bbt2eig B (U, Σ)
ell chol2eig L (U, Σ)
ellu chol2eig L (U, Σ)
ell eig2chol (U, Σ) L
ell full2eig A (U, Σ)
ell full2low A Ā
ell low2chol Ā L

4.1 Linear transformation

We consider the linear transformation

y = MT (x − b), (19)

where b is a specified vector and M is a specified non-singular n × n matrix.

The inverse transformation is

x = b + M−T y. (20)

If the ellipsoid E is defined by

E ≡ {x | kBT (x − c)k ≤ 1}, (21)

then it is readily shown, from Eq.(20), that it is also given by

T
E = {y | kB̂ (y − ĉ)k ≤ 1}, (22)

with
B̂ ≡ M−1 B, (23)
and
ĉ ≡ MT (c − b). (24)

6
 Table 2: Tasks performed by principal routines in ELL LIB
Routine Section Task performed
ell rad lower 5 Determine the smallest
principal semi-axis of E
ell rad upper 5 Determine the largest
ellu rad upper principal semi-axis of E
ell radii upper 5 Determine the smallest and largest
ellu radii upper principal semi-axes of E
ell pt in 6 Determine whether
p is covered by E
ell pt dist 7 Determine the relative distance from
p to the boundary of E
ell pt near far 8 Determine the point in E which is
ellu pt near far 9 nearest to or furthest from p
ell pt modify 10.2 Determine EV , the modification
of E of least change of
content whose boundary
intersects with p
ell pt shrink 11 Shrink E based on p
ell pt hyper - Determine a separating hyperplane
between E and p
ell pts uncover 14 Determine an ellipsoid (of
large content) which does
not cover any given point
ell line proj 12 Project E onto a given line
ell aff pr 13 Project E onto a given affine space
ell pair shrink 17 For concentric E1 and E2 ,
shrink E2 (if necessary)
so that it is covered by E1
ell pair separate 15 Determine if E1 and E2
intersect; and, if they
do not, determine a
separating hyperplane
ell pair cover query 16 Determine wether E1 covers E2
ell pair cover 18 Determine an ellipsoid
(of small content) which
covers E1 and E2

7
4.2 Quadratic minimization
Several of the algorithms described below depend on the solution to the fol-
lowing problem: determine a vector x which minimizes the quadratic function

g(x) ≡ 21 xT Ax + bT x, (25)
subject to
xT x ≤ δ 2 , (26)
where δ, b and A are a given scalar, vector, and symmetric matrix, re-
spectively. Such problems are efficiently solved by the routine dgqt from
MINPACK-2 (see Averick et al. 1993).

4.3 Householder matrix

Given a vector p with p ≡ kpk > 0, the corresponding Householder matrix
H(p) is defined by
H = I − 2vvT , (27)
where v is the unit vector
1
v1 = [(1 + |p1 |)/(2p)] 2 ,
sign (p1 )pi
vi = , for i ≥ 2. (28)
2v1 p

It is readily shown that H has the following properties (not all independent):

1. H is symmetric: HT = H.

2. H is orthogonal: HT H = I.

3. The first column of H is parallel (or anti-parallel) to p.

4. All columns of H except the first are orthogonal to p.

5. H maps p to the first axis: Hp = pe1 .

Given p, the routine ell house returns v.

8
4.4 Rank-one modification
Let E be an ellipsoid centered at the origin (i.e., c = 0), given in terms of
the positive symmetric definite matrix A, which has SVD A = UΣ2 UT , and
Cholesky decomposition A = LLT . For a given vector w and scalar ρ, let F
be the rank-one modification of A:
F = A + ρwwT . (29)
We require F to be positive symmetric definite, which in turn requires ρ
to exceed a critical value ρ0 (ρ0 < 0) above which all eigenvalues of F are
positive (see item 7 below). Then, the modified ellipsoid E 0 is defined by
E 0 ≡ {x | xT Fx ≤ 1}. (30)
The following results are readily obtained:
1. For ρ ≥ 0, xT Fx is greater than or equal to xT Ax, and hence E covers
E 0.
2. For ρ ≤ 0, xT Fx is less than or equal to xT Ax, and hence E 0 covers
E.
3. The eigenvalues of F are interlaced with those of A (see Golub and
Van Loan 1996, Sec. 8.5.3). Hence the lengths of the principal axes of
E and E 0 are also interlaced.
4. With z ≡ ΣUT x, we have xT Ax = zT z, so that E is the unit ball in
z-space. Correspondingly, E 0 is given by
E 0 = {z | zT (I + ρw̃w̃T )z ≤ 1}. (31)
where
w̃ = Σ−1 UT w. (32)
− 12
Thus in z-space, one principal semi-axis of E 0 is (1 + ρ|w̃|2 ) w̃/|w̃|,
and the others are orthogonal unit vectors.
5. Similarly, with y ≡ LT x, we have xT Ax = yT y, so that E is the unit
ball in y-space. Correspondingly, E 0 is given by
E 0 = {y | yT (I + ρŵŵT )y ≤ 1}. (33)
where
ŵ = L−1 w. (34)

9
 6. The rank-one modifications to the identity appearing in Eq.(31) has
the symmetric square root

I + ρw̃w̃T = G2 = (I + αw̃w̃T )(I + αw̃w̃T ) (35)

where α and ρ are related by

ρ = 2α + α2 |w̃|2 , (36)

1
α = (±{1 + ρ|w̃|2 } 2 − 1)/|w̃|2 . (37)
The matrix in Eq.(33) has a similar square root, with the same value
of α, since |ŵ| = |w̃|.

7. The critical value of ρ is

ρ0 = −1/|w̃|2 , (38)

corresponding to the smallest value of ρ for which α is real.

5 Smallest and largest principal semi-axes of

E
As sketched in Fig. 1, given an ellipsoid E, the task is to determine its
smallest and largest principal semi-axes, or equivalently, the radii of the
inscribed and circumscribed hyper-spheres, rin and rout , respectively. This
task is accomplished by the routines ell rad lower and ell rad upper.
Considering the ellipsoid E, Eq.(16), let the SVD of the Cholesky factor
L be
L = UΣVT , (39)
with σ1 ≥ σ2 ≥ . . . ≥ σn being the components of the diagonal matrix of
singular values Σ. Then the matrix A, Eq.(15), is

A = LLT = UΣ2 UT = UΛUT . (40)

Evidently the columns of U are eigenvectors of A, and the diagonal matrix

Λ consists of the eigenvalues of A with components

λi = σi2 . (41)

10
 E

rin

rout

Figure 1: Sketch of an ellipsoid E showing the radii, rin and rout of the inscribed
and circumscribed hyper-spheres.

− 12
The principal semi-axes are given by ri = λi = σi−1 . Thus the smallest
principle semi-axis is
1
rin = [max(λi )]− 2 , (42)
and the largest is
1
rout = [min(λi )]− 2 . (43)
The most stable way to compute rin is as rin = 1/σ1 , where σ1 is obtained
from the SVD of L, and similarly for rout . The routine ell radii determines
both rin and rout using the SVD.
The routines ell rad lower and ell rad upper determine rin and rout at sig-
nificantly lower computational cost, but with less accuracy. Given L, the
routine ell rad lower determines rin via Eq.(42), using the LAPACK routine
dsyevx to compute the largest eigenvalue of A.
The routine ell rad upper determines rout via quadratic minimization. Since
rout is the furthest distance from the center to any point on the ellipsoid we
2
have that rout is the maximum of rT r subject to

rT Ar ≤ 1. (44)

By defining y = LT r, this can be put in the standard form of a quadratic

2
minimization problem: −rout is the minimum of

−yT L−1 L−T y, (45)

11
subject to
yT y ≤ 1. (46)

6 Is the point x covered by E?

Given a point x and an ellipsoid E (in terms of c and L), the task is to
determine whether E covers x. This is readily determined (by the routine
ell pt in) through the definition of E given by Eq.(17). That is, E covers x
if the quantity
s ≡ kLT (x − c)k, (47)
is less than or equal to unity.

7 Relative distance to the boundary of E

Given an ellipsoid E (in terms of c and L) and a point p (p 6= c), let b be
the intersection of the ray c-p with the boundary of E. The relative distance
s to the boundary is defined by:

s ≡ |b − c|/|p − c|. (48)

Now the boundary point satisfies

kLT (b − c)k = 1, (49)

and we have
b − c = s (p − c). (50)
Hence, s is determined as

s = kLT (p − c)k−1 . (51)

The routine ell pt dist determines s.

It may be noted that the three cases s < 1, s = 1 and s > 1 correspond,
respectively, to: p not being covered by E; p being on the boundary of E;
and, p being covered by E.

12
 f

Figure 2: The points n and f in the ellipsoid E which are nearest and furthest,
respectively, to the given point p.

8 Nearest point in E to a given point

We consider the ellipsoid E Eq.(16) centered at c, and a given point p. The
task (performed by the routine ell pt near far)is to find the point x in E
closest to p (see Fig.2). Thus x satisfies
(x − c)T LLT (x − c) ≤ 1, (52)
and minimizes
s2 ≡ (x − p)T (x − p). (53)
Evidently, if p is contained in E then x = p and s = 0. Otherwise x is on
the boundary of E.
The above equations are readily transformed into the standard quadratic
minimization problem. Let y be defined by
y ≡ LT (x − c), (54)
so that Eq.(52) becomes
yT y ≤ 1. (55)
Equation (54) can be inverted to yield
x = c + L−T y, (56)
so that Eq.(53) becomes
s2 = (c + L−T y − p)T (c + L−T y − p)
= yT L−1 L−T y + 2(c − p)T y + (c − p)T (c − p). (57)

13
Thus y is obtained via dgqt by minimizing
χ ≡ 12 yT L−1 L−T y + (c − p)T y, (58)
subject to kyk ≤ 1; and then x is obtained from Eq.(56).

9 Furthest point in E to a given point

This task is essentially the same as that considered in the previous section,
except that the furthest point in E, x, maximizes s2 ≡ kx − pk2 , Eq.(53)
(rather than minimizing s2 ). Thus the same algorithm is used, but the
quantity minimized is −χ, Eq.(58) (rather than χ). Again, this is performed
by the routine ell pt near far.
For all given points p, the corresponding furthest point x lies on the
boundary of the ellipsoid E (see Fig.2).

10 Minimum-volume ellipsoid covering E and

p
Given an ellipsoid E and a point p, are seek the ellipsoid E 0 (concentric with
E) of minimum volume which covers both E and p. In the first subsection
we describe an algorithm based on the geometry of the problem. In the
subsequent subsection a simpler algorithm is given, which is implemented in
the routine ell pt modify. Then the behavior of the algorithm is examined.

10.1 Householder matrix algorithm

As sketched in Fig. 3(a), consider a given point p lying outside an ellipsoid
E. A modified concentric ellipsoid E 0 is sought which has minimum volume
subject to:
1) E 0 intersects p
2) E 0 covers E.
This is achieved by
1) performing the linear transformation which transforms E to the unit
hypersphere, and p to a point p̂ on the first axis, see Fig. 3 (b)

14
 (a)



p (b)
c
E
 
 



p̂

(c) (d)

p

c


E0
 


p̂

Figure 3: Sketch of: (a) the given ellipsoid E and point p; (b) the transformation
of E to the unit ball, and p to a point on the first axis; (c) the transformed ellipsoid
E 0 ; and (d) the grown ellipsoid E 0 .

2) extending the hypersphere in the first direction to form an ellipsoid

with p̂ on its boundary, see Fig. 3 (c)

3) inverting the initial transformation to obtain E 0 , see Fig. 3 (d).

The ellipsoid E (Eq.(17)) is defined by

kLT rk = kLT (x − c)k ≤ 1. (59)

Defining the transformed variable y by

y = LT (x − c), (60)

evidently the ellipsoid in y-space is the unit hypersphere

kyk ≤ 1. (61)

The same transformation applied to p yield

p̃ = LT (p − c). (62)

15
 There is an orthogonal matrix Q which transforms p̃ to a point on the
1-axis; that is,
p̂ = QT p̃ = QT LT (p − c), (63)
where
p̂i = p̃δi1 , (64)
and p̃ = kp̃k. It is readily shown that Q is simply the Householder matrix
Q = H(p̃). The same transformation applied to E yields the unit ball

zT z ≤ 1, (65)

where
z = QT LT (x − c). (66)
This ball and the point p̂ (p̂i ≡ p̃δi1 ) are shown in Fig. 3 (b). The next step
is to define the modified ellipse E 0 in z-space, as shown in Fig. 3 (c). This
ellipsoid is
zT Λ2 z ≤ 1, (67)
where
Λ2ij = δij + δi1 δj1 (p̃−2 − 1). (68)
Transforming back to the original space, we obtain the equation for E 0
(see Fig. 3 (d)):
(x − c)T LQΛ2 QT LT (x − c) ≤ 1, (69)
or
(x − c)T L0 L0T (x − c) ≤ 1. (70)
Thus the modified ellipsoid E 0 has center c and Cholesky matrix L0 given by

L0 L0T = LQΛ2 QT LT
= (LQΛ)(LQΛ)T . (71)

In summary, given the ellipsoid E (in terms of c and L) and the point p,
then p̃ is determined by Eq.(62), v by Eq.(28), Q by Eq.(27), Λ2 by Eq.(68),
and finally the Cholesky matrix of the modified ellipsoid E 0 is determined by
Eq.(71).

16
10.2 Rank-one modification algorithm
Following from Eqs.(61) and (62), in y-space, the modified ellipsoid E 0 is the
unit ball extended to intersect the point p̃. Thus, based on the results of
Sec. 4.4, we have
E 0 = {y | yT G2 y ≤ 1}, (72)
where
G = I + γ p̃p̃T , (73)
and γ (determined by the condition p̃T G2 p̃ = 1) is
à !
1 1
γ= −1 . (74)
|p̃| |p̃|2

Thus, in place of Eq.(71), the Cholesky matrix L0 can be obtained as

L0 L0T = (LG)(LG)T . (75)

This algorithm is implemented in the routine ell pt modify.

10.3 Behavior
Test are reported for the minimum-volume growing algorithm. The tests are
in two dimensions and are based on an initial ellipsoid E (centered at the
origin and aligned with the coordinate axes). The length of the major semi-
axis is rmajor = 1, and the minor semi-axis is rminor . The grow point p is not
covered by E, and the vector p is at an angle θ to the x1 axis. An example
of the grown ellipse Ẽ given by the minimum-volume algorithm is shown in
Fig. 4.
The lengths of the principal semi-axes of the grown ellipse Ẽ are denoted
by Rmajor and Rminor . A figure of demerit F of the growth operation is
defined by:
F ≡ Rmajor / max(rmajor , |p|) ≥ 1. (76)
Tests suggest the following behavior:
1. For given E and θ, the maximum of F occurs for |p| = rmajor , i.e., for
p being on the bounding circle.

2. As illustrated in Fig. 5, for given E, F has a unique maximum (Fmax )

at θ = θmax between θ = 0 and θ = π/2.

17
 Minimum−volume growing for grow point at 70 degrees

0.5

−0.5

−1

−1.5 −1 −0.5 0 0.5 1 1.5

Figure 4: Illustration of the minimum-volume algorithm. The initial ellipse E

(solid blue line); its bounding circle (dashed blue line); the grow point p (magenta);
and the grown ellipse (red line) Ẽ: rminor = 0.1, |p| = rmajor = 1, θ = 70o .

3. √
As rminor decreases, θmax tends towards π/2, and Fmax tends towards
2.

With some justification,

√ it is speculated that even in higher dimensions
F is bounded above by 2.

11 Shrink E based on a given point

Given an ellipsoid E and a point p, the problem considered is to generate a
concentric ellipsoid E 0 of smaller volume which covers p. There is no unique
solution to this problem. In the following three subsections we describe three
algorithms which yield different solutions for E 0 , denoted by EV , EN and EC ,
respectively.
The third (which is based on the first two) is preferred in application to
ISAT. The three algorithms are implemented in the routine ell pt shrink with
the parameter k ell set to 1, 2 and 3, respectively.

18
 Amplification of the major axis against the angle of the grow point
1.4

1.35

1.3

1.25
Rmajor/rmajor

1.2

1.15

1.1

1.05

0.95
0 10 20 30 40 50 60 70 80 90
θ

Figure 5: The demerit measure F as a function of the grow-point angle θ: rminor =

0.1, |p| = rmajor = 1.

11.1 Maximum-volume algorithm

11.1.1 Algorithm
The algorithm described in section 10 to grow on ellipsoid E based on a point
p lying outside of E can also be applied when the point p lies inside E.
The resulting modified ellipsoid EV has maximal volume subject to:

1) p is on the boundary of EV

2) EV is covered by E.

11.1.2 Behavior
We consider the same 2D case as for the minimum-volume growing algorithm,
except that here the shrink point p is covered by the ellipse E. The maximum-
volume shrinking algorithm determines the ellipse EV of maximum volume
which is covered by E and has p on its boundary. This is illustrated in Fig. 6.
We define a figure of merit

F ≡ Rminor / min(rminor , |p|) ≤ 1, (77)

19
Figure 6: The maximum-volume shrinking algorithm. The initial ellipse E (solid
blue line); its inscribed circle (green line); the grow point p (magenta); and the
maximum-volume shrunk ellipse EV (red line): rminor = 0.4, rmajor = 1, θ = 70o ,
|p| = 0.48.

Figure 7: The maximum-volume shrinking algorithm. The initial ellipse E (solid

blue line); its inscribed circle (green line); the grow point p (magenta); and the
maximum-volume shrunk ellipse EV (red line): rminor = 0.1, |p| = rmajor = 1,
θ = θmin = 84.3o .

where rminor and Rminor are the lengths of the minor semi-axes before and
after shrinking. Based on empirical testing we draw the following conclusions.
1. For given E and θ, the minimum (Fmin ) of F occurs for p = rminor ,
i.e., for p lying on the inscribed circle.
2. As rminor /rmajor decreases, the values of θ at which F is minimum (i.e.,
F (θmin ) = Fmin ) tends to π/2 − rminor /rmajor , and Fmin tends to zero
as 2rminor /rmajor .
Figure 7 shows the case rminor /rmajor = 0.1, θ = θmin = 84.3o for which
Fmin = 0.1962.

11.2 Near-content algorithm

As shown above, the maximum-volume shrinking algorithm can lead to very
small figures of merit F , Eq.(77). As may be seen from Fig. 7, the shrunk

20
ellipsoid can exclude a substantial portion of the original ellipsoid which is
closer to the center than the shrink point. Here we describe the alternative
“near-content shrinking algorithm” which yields values of F of one, or close
to one. However, unlike the maximum-volume algorithm, the result depends
on the metric of the space. It is implemented in the routine ell pt shrink (for
the parameter value k ell= 2).

11.2.1 Algorithm
We first describe the algorithm and then give its partial justification.
The initial ellipsoid E centered at c is given in terms of the matrix A with
SVD A = UΣ2 UT . We transform to principal axes (y-space) by defining

y = UT x, (78)

and
p̂ = UT (p − c). (79)
In y-space, the modified ellipsoid EN is defined as the rank-one modification
to E:
EN = {y | yT Fy ≤ 1}, (80)

F = Σ2 + ρwwT , (81)

w = Dp̂, (82)
where the diagonal matrix D is defined by

Dii = max(0, 1/|p̂|2 − Σ2ii ), (83)

and the positive scalar ρ is determined by the intersection condition

p̂T Fp̂ = 1. (84)

We make the following observations about the algorithm:

1. The quantity 1/|p̂|2 − Σ2ii (appearing in Eq.(83)) is positive if, and only
if, the length of the i-th principal semi-axis Σ−1
ii is greater than |p̂|.

21
 2. Since E covers p̂, p̂T Dp̂ is strictly positive, and hence a positive value
of ρ exists which satisfies Eq.(84), namely

ρ = (1 − p̂T Σ2 p̂)/(p̂T Dp̂). (85)

3. If p̂ lies inside the ball of radius Σ−1 2 2

11 , then Dii = 1/|p̂| − Σii , p̂ is an
eigenvector of F, and hence p̂ is the smallest principal semi-axis of EN .
This is a necessary condition for EN to be a maximum nearest content
ellipsoid.

4. If for some i (1 ≤ i < n), p̂ lies between the balls of radius Σ−1 ii
and Σ−1(i+1)(i+1) , then wj = 0 for j ≤ i, and as a consequence the first
i principal axes of EN are the same as those of E. This again is a
necessary condition for EN to be a maximum nearest content ellipsoid.

11.2.2 Behavior
The behavior of the near-content algorithm is illustrated for a 2D case in
Fig. 8; and it is compared to the maximum-volume algorithm in Fig. 9. There
is a striking difference between the two methods, especially for relatively
small |p|.

11.3 Conservative algorithm

As may be seen from the last pane in Fig.9, the shrunk ellipsoids (denoted
by EV and EN ) generated by the maximum-volume and near-content algo-
rithms can be quite different. In particular each excludes substantial regions
included by the other. Here we describe a third algorithm which is “con-
servative” in the sense that the shrunk ellipsoid generated, EC , includes all
points in EV and EN . Specifically, EC is defined as the ellipsoid of minimum
volume which covers both EV and EN . This algorithm is implemented in the
routine ell pt shrink (for the parameter value k ell= 3).
The behavior of the conservative algorithm is illustrated in Fig. 10. It
has the following properties:
1. EC covers p. Generally, p is in the interior of EC .

2. EC covers both EV and EN .

3. In general EC is not covered by E.

22
 Max. close content: |p|/rmin=1.5 Max. close content: |p|/rmin=1.01

0.5 0.5

0 0

−0.5 −0.5

−1 −0.5 0 0.5 1 −1 −0.5 0 0.5 1

Max. close content: |p|/rmin=0.99 Max. close content: |p|/rmin=0.5

0.5 0.5

0 0

−0.5 −0.5

−1 −0.5 0 0.5 1 −1 −0.5 0 0.5 1

Figure 8: The near-content shrinking algorithm. The initial ellipse E (solid blue
line); its inscribed circle (light blue line); the grow point p (magenta); and the
maximum near content shrunk ellipse EN (red line): rmajor = 1, rminor = 0.4,
θ = 70o .

4. The largest principal axis of EC can exceed that of E. Tests suggest

that this ratio of principal axes is seldom greater than 1.3.

5. The volume of EC is no greater than that of E, and in general is less.

(This follows from the fact that both E and EC cover EV and EN , but
EC is, by definition, of minimum volume.)

6. If the algorithm is applied a second time to EC based on the original

point p, the results is (in general) a different ellipsoid. (This is in
contrast to the maximum-volume and near-content algorithms in which
p intersects the boundaries of EV and EN and hence a re-application
of the algorithm has no effect.)

7. Because the algorithm involves EN , the result EC depends on the metric

of the space.

23
 Max. close content: |p|/rmin=1.6 Max. close content: |p|/rmin=1.2

0.5 0.5

0 0

−0.5 −0.5

−1 −0.5 0 0.5 1 −1 −0.5 0 0.5 1

Max. close content: |p|/rmin=0.6 Max. close content: |p|/rmin=0.2

0.5 0.5

0 0

−0.5 −0.5

−1 −0.5 0 0.5 1 −1 −0.5 0 0.5 1

Figure 9: Comparison of the near-content and maximum-volume shrinking algo-

rithms. The initial ellipse E (solid blue line); its inscribed circle (light blue line);
the grow point p (magenta); the near content shrunk ellipse EN (red line); and
the maximum-volume shrunk ellipse EV (green line): rmajor = 1, rminor = 0.4,
θ = 70o .

8. Tests reveal that the figure of merit F Eq.(77) is never less than unity.

The algorithm to generate EC consists of reducing the problem to two

dimensions, and then solving a 2D problem. These two parts of the algorithm
are described in the following two subsections.

11.3.1 Algorithm: reduction to 2D

Following the development in Sec. 11.2, we transform to the principal axes
of E (y-space), defining y, p̂ and w by Eq.(78), Eq.(79) and Eq.(82). Then
EN is defined by
EN = {y | yT Fy ≤ 1}, (86)
with
F = Σ2 + ρwwT . (87)

24
 Max. close content: |p|/rmin=1.6 Max. close content: |p|/rmin=1.2

0.5 0.5

0 0

−0.5 −0.5

−1 −0.5 0 0.5 1 −1 −0.5 0 0.5 1

Max. close content: |p|/rmin=0.6 Max. close content: |p|/rmin=0.2

0.5 0.5

0 0

−0.5 −0.5

−1 −0.5 0 0.5 1 −1 −0.5 0 0.5 1

Figure 10: Conservative shrinking algorithm. The initial ellipse E (solid blue line);
the grow point p (magenta); the maximum-volume shrunk ellipse, EV , (green line);
the near-content shrunk ellipse, EN , (red line); and the conservative shrunk ellipse,
EC , (magenta): rmajor = 1, rminor = 0.4, θ = 70o .

We now transform to z-space in which E is the unit ball:

z = Σy = ΣUT x. (88)

In z-space the definition of EN , Eq.(86), becomes

EN = {z | zT (I + ρw
e w)z
e ≤ 1}, (89)

with
e = Σ−1 w.
w (90)
And the maximum volume shrunk ellipsoid is

EV = {z | zT (I + γ p
epe T )z ≤ 1}, (91)

where p
e is the transform of p

25
 ž2

1 EC

EN p̌
−1 χ 1 ž1

EV

−1

Figure 11: Sketch of E, EN and EV in the ž1 -ž2 plane (in which E is the unit
disc) and of EC which covers EN and EV . The boundaries of both EN and EV
intersect the grow point p̌.

e = Σp̂ = ΣUT (p − c),

p (92)

(see Eq.(72)). The scalars ρ and γ are determined by the condition that the
boundaries of EN and EV intersect p e (in z-space).
Note that EN and EV are completely determined by two vectors, w e and
p
e . If these are co-linear, then EN = EV = EC . Otherwise we perform a
rotation
ž = QT z, w̌ = QT w,
e p̌ = QT p
e, (93)
such that w̌ is in the ž1 -direction, and p̌ is in the ž1 -ž2 plane. The appropriate
orthogonal matrix Q is obtained from the QR decomposition

QR = [w
e p
e ]. (94)

26
 Figure 11 is a sketch of the intersection of the ellipsoids with the ž1 -ž2
plane. Note that in the other directions the principal semi-axes are unit
vectors aligned with the coordinate axes. The minimum volume ellipsoid EC
covering EN and EV is also shown in the figure. Its intersection with the
ž1 -ž2 plane is given by the ellipse
" #
ž1
EC0 = {(ž1 , ž2 ) | kL̄ T
k ≤ 1}, (95)
ž2

where L̄ is a 2×2 Cholesky matrix which is determined in the next subsection.

Thus the ellipsoid EC is given by
T
EC = {ž | kĽ žk ≤ 1}, (96)

where the n × n Cholesky matrix Ľ is

" #
L̄ 0
Ľ = . (97)
0 I

Transforming Eq.(96) back to the original x-space, we obtain

EC = {x | kLTC (x − c)|| ≤ 1}, (98)

where the Cholesky matrix LC is obtained from

T
AC = LC LC = BC BTC , (99)

BC = UΣQĽ. (100)

11.3.2 Algorithm: solution in 2D

The problem to be solved is the determination of the 2 × 2 Cholesky matrix
L̄ defining the covering ellipse EC in the ž1 -ž2 plane (see Fig. 11).
The scalars ρ and γ in Eq.(89) and Eq.(91) are determined by the inter-
section condition (of ∂EN and ∂EV with p̌) to be

ρ = (1 − |p̌|2 )/p̌21 , (101)

27
 γ = (1 − p̌|2 )/|p̌|4 , (102)
and the intersection of EN with the ž1 axis is at ž1 = χ

|p̌1 |
χ= q . (103)
1 − p̌22

A transformation to ζ-space is performed

" # " # " #" #
ζ1 ž1 1/χ 0 ž1
=C = , (104)
ζ2 ž2 0 1 ž2

consisting of a stretching in the ž1 direction so that EN transforms to the

unit disc.
With this transformation EV becomes
f ≤ 1},
EV = {ζ | ζ T Aζ (105)

with " # " #

f= a c χ2 (1 + γ p̌21 ) γχp̌1 p̌2
A = . (106)
c b γχp̌1 p̌2 1 + γ p̌22
f is
The SVD of A
f=U
A fΣe 2U
fT , (107)
with " #
f= cos θ sin θ
U , (108)
− sin θ cos θ

θ = 12 tan−1 (−2c/(a − b)), (109)

e 2 = a cos2 θ − 2c sin θ cos θ + b sin2 θ,

Σ (110)
11

e 2 = a sin2 θ + 2c sin θ cos θ + b cos2 θ.

Σ (111)
22

Now EC is the ellipse which covers EN (which is the unit disc) and EV
which has principal semi-axes whose directions are given by the columns of
f and whose lengths are Σ
U e −1 . (In general, one of Σ
e −1 and Σ e and Σe is
11 22 11 22
less than unity and one greater than unity.) Thus the covering ellipse EC
has the same principal directions

28
 fΣ
EC = {ζ | ζ T U b 2U
fT ζ ≤ 1}, (112)

and eigenvalues
b = min(1, Σ
Σ e ), (113)
11 11

b = min(1, Σ
Σ e ). (114)
22 22

Transforming back to ž1 -ž2 we obtain the required result that EC is given
by Eq.(95), where the 2 × 2 Cholesky matrix L̄ is obtained as
T fΣ
L̄L̄ = CU fT C.
b 2U (115)

12 Orthogonal projection of E onto a given

line
We consider a given line L, parameterized by s, defined by

L ≡ {x | x = x0 + sv}, (116)

where x0 is a given point and v is a given non-zero vector, see Fig. 12. Given
any point x in the space, its orthogonal projection onto L corresponds to

vT (x − x0 )
s= . (117)
vT v
Now the given ellipsoid E is given by

E = {x | x = c + L−T y, ||y|| ≤ 1}. (118)

Thus the projection of points in E correspond to values of s

vT (L−T y + c − x0 )
s =
vT v
= s0 + wT y, for kyk ≤ 1, (119)

where
vT (c − x0 )
s0 ≡ , (120)
vT v

29
 s
s+

c
s0

L
E
s−
v
x0

Figure 12: The orthogonal projection of the ellipsoid E onto the line L is the
interval [s− , s+ ].

and
L−1 v
w≡ . (121)
vT v
Given the condition kyk ≤ 1, it is evident from Eq.(119) that the orthogonal
projection of E onto L corresponds to the interval [s , s+ ] in s, with

s± ≡ s0 ± |w|. (122)

This is sketched in Fig. 12. This method is implemented in the routine

ell line proj.

13 Orthogonal projection of E onto an affine

space
We consider an ellipsoid E (in <n ), an affine space A (in <m ), for 1 ≤ m ≤ n,
and the orthogonal projection P (E) of E onto A.
The ellipsoid E is given by

E ≡ {x | x = c + L−T u, kuk ≤ 1}. (123)

The affine space A is given by

30
 A ≡ {x | x = d + Tt}, (124)
where d is a given m-vector, T is a given n × m orthogonal matrix, and t
is a vector of m parameters. The orthogonal projection of a general point x
onto A is
P (x) = d + TTT (x − d). (125)
We thus obtain
P (E) = {x | x = d + TTT (c − d + L−T u), kuk ≤ 1}. (126)
Now let the SVD of TT L−T be
TT L−T = U[Σ 0]VT . (127)
Then
TT L−T u = UΣw̃, (128)
where w̃ denotes the first m elements of
w ≡ VT u. (129)
Note that kuk ≤ 1 implies kw̃k ≤ 1. Thus we obtain
P (E) = {x | x = d + T(c̃ + UΣw̃), kw̃k ≤ 1}, (130)
where
c̃ ≡ TT (c − d). (131)
The above equation for P (E) is for an m-dimensional ellipsoid in A. It
can be put in standard form by defining B by

B−T = UΣ, (132)

and then L̃ as
T
L̃L̃ = BBT , (133)
so that we can write
−T
P (E) = {x | x = d + T(c̃ + L̃ w̃), kw̃k ≤ 1}. (134)
The Cholesky matrix L̃ can be computed from the LQ decomposition
UΣ−1 = L̃Q. (135)
This method is implemented in the routine ell aff pr.

31
14 Generate an ellipsoid which does not cover
any specified points
We are given a point c, a set of P points p(j) (j = 1 : P ) and a positive length
rmax . The problem is to generate an ellipsoid E which

1. is centered at c

2. has principal semi-axes no larger than rmax

3. does not cover any of the P points

4. and is “as large as possible” (in an undefined sense).

The algorithm used to solve this problem is implemented in the routine

ell pts uncover. It involves a user-specified parameter θ (0 < θ ≤ 1) which
affects the shape of the resulting ellipsoid, E.
The algorithm has two phases. In the first phase there are n stages which
generate a succession of ellipsoids E1 , E2 , . . ., En . In the second phase, E is
formed by shrinking En uniformly and minimally so that none of the points
is covered.
In the first phase, a principle axis is determined on each stage. The ellip-
soid Ek is determined on the kth stage, and it has the following properties:

1. Ek is centered at c

2. for 1 ≤ ` < k, the `th principle axis of Ek is the same as that of E`

(previously determined on stage `)

3. the kth principle axis (of half-length rk ≤ rmax ) is determined on the

kth stage

4. for k < ` ≤ n, the `th principle axis of Ek is of half-length rk .

Note that E1 is a ball of radius r1 .

An orthonormal basis is developed with basis vectors e1 , e2 , ..., en . On
the kth stage e` (` ≥ k) is modified, but subsequently ek is not altered. At
the end of the kth stage, the basis vectors are principal axes of Ek . The
vectors y(j) (j = 1 : P ) store the coordinates of the points (relative to c) in
the current basis. For the jth particle we define

32
 n
X (j)
hj = (yi /ri )2 . (136)
i=1

At the end of the kth stage (since r` = rk for ` ≥ k) we can decompose hj as

hj = fj + gj /rk2 , (137)

where
k−1
X (j)
fj = (yi /ri )2 , (138)
i=1

and n
X (j)
gj = (yi )2 . (139)
i=1

The ellipsoid Ek does not cover the point j if hj is greater than unity. This
condition (hj > 1) can be re-expressed as

gj /rk2 > (1 − fj ). (140)

Points with fj > 1 cannot be covered by Ek regardless of how large rk is.

Such points are “excluded.” Points with

θ2 < fj ≤ 1, (141)

are “partially excluded.” Such points cannot be covered by Ek shrunk by a

factor of θ. The remaining points (i.e., with fj ≤ θ2 ) are “included.”
We define r̂k2 to be the minimum value over the included points of gj /(1 −
fj ), and denote by ĵ the index of a point that achieves this minimum. The
significance of r̂k is that if rk is set to r̂k , then the point ĵ is on the boundary
of Ek , but no points are in the interior of Ek . If r̂k is greater than rmax (or if
there are no included points), then we get r` = rmax for all ` ≥ k, and omit
the remaining stages of the first phase. Otherwise rk is set to r̂k , and the
(ĵ)
basis vectors e` (k ≤ ` ≤ n) are re-defined so that y` = 0 for ` > k. In this
way, in subsequent stages, the ellipsoid can expand in directions orthogonal
to y(ĵ) , with y(ĵ) remaining on the boundary.
At the end of the first phase there are no included points, and En does
not cover any excluded points. However En may cover one or more partially
excluded points. Consequently, the final result E is obtained by shrinking En
uniformly, as little as possible so that none of the partially included points
is covered.

33
 (a) (b) E10

O
E1
v
H0
xh y1
H yh

E2 y2

E20

Figure 13: (a) Sketch of the non-intersecting ellipsoids E1 and E2 and a separating
hyperplane H (b) Corresponding sketch in the transformed space in which E10 is
the unit ball.

15 Separating hyperplane of two ellipsoids

Given two ellipsoids, E1 and E2 , the task is to determine if they intersect;
and, if they do not intersect, to determine a separating hyperplane, H. This
is performed by the routine ell pair separate.
The ellipsoid E1 is given by

E1 ≡ {x | kLT1 (x − c1 )k ≤ 1}, (142)

or, equivalently,

E1 ≡ {x | kyk ≤ 1, y ≡ LT1 (x − c1 )}.

With the transformation

y ≡ LT1 (x − c1 ), (143)
E1 is transformed to the unit ball at the origin (denoted by E10 ) and E2 is
transformed to E20 , see Fig. 13. The point y2 in E20 which is closest to the
origin is determined by the closest point algorithm (see Sec. 8). If ky2 k is
less than unity, then E1 and E2 intersect.
For the case in which E1 and E2 do not intersect, we define

y1 = v = y2 /ky2 k, (144)

34
so that y1 and y2 are the pair of closest points in E10 and E20 . Note that v is
a unit vector. Then we define

yh ≡ 21 (y1 + y2 ), (145)

and a separating hyperplane is defined by

H 0 ≡ {y | vT (y − yh ) = 0}. (146)

Inverting the transformation Eq.(143), we obtain the separating hyper-

plane in the original space:

H ≡ {x | uT (x − xh ) = 0}, (147)

where
L1 v
u= , (148)
kL1 vk
and
xh = c1 + L−T
1 yh . (149)

For a hyperplane H given by Eq.(147) (for some u and xh ), we define the

quality q as follows. Let x1 be the point in E1 closest to H, and similarly let
x2 be the point in E2 closest to H. The quality q is defined as

uT (x2 − x1 )
q≡ ≤ 1. (150)
|x2 − x1 |
This is the distance between the supporting hyperplanes at x1 and x2 relative
to their separation. The maximum possible distance between the supporting
hyperplanes is achieved for q = 1, and this occurs when x1 and x2 are the
mutually closest points in E1 and E2 .
In the routine ell pair separate there is an option to iteratively improve
the quality of the hyperplane. Initially x1 and x2 are set to the points in E1
and E2 corresponding to y1 and y2 (in E10 and E20 ). Then, successively, x1
is replaced by the closest point in E1 to x2 ; and then x2 is replaced by the
closest point in E2 to x1 . The hyperplane is then taken as the perpendicular
bisector of the line x1 − x2 . It is not guaranteed that this hyperplane is
separating, nor that the quality increases with the iterations, although it
generally does. Consequently, H is taken as the hyperplane with the greatest
quality encountered initially or during the iterations.

35
 (Note that the separating hyperplane of quality q = 1 could alternatively
be obtained by solving the quadratic programming problem, of determining
the mutually closest points x1 and x2 in E1 and E2 . This has not been
implemented.)

16 Pair covering query

Given a pair of ellipsoids E1 and E2 (centered at c1 and c2 and with Cholesky
factors L1 and L2 ), the problem is to determine whether E1 covers E2 . This
query is answered by the routine ell pair cover query by the following algo-
rithm.
With the same transformation as used in Sec. 15, E1 is transformed to
the unit ball, at the origin, and E2 to E20 (see Eq.(143) and Fig.13). The
algorithm described in Sec. 9 is then used to determine the distance s from
the origin to the furthest point in E20 . The ellipsoid E1 covers E2 if and only
if s is less than or equal to unity.

17 Shrink ellipsoid so that it is covered by a

concentric ellipsoid
Given two concentric ellipsoids, E1 , and E2 , the task is to form the maximal-
volume ellipsoid, E, which is covered by both E1 and E2 . This is performed
by the routine ell pair shrink.
Clearly E is concentric with E1 and E2 , and so, without loss of generality,
we take the origin at the mutual center. Then, E1 is given by

E1 ≡ {x | kLT1 xk ≤ 1}, (151)

and similarly for E2 and E (in terms of L2 and L, respectively).

The transformation
y = LT1 x, (152)
maps E1 to the unit ball, E10 , and it maps E2 to the ellipsoid E20 with Cholesky
factor
L02 = L−1
1 L2 , (153)
see Fig. 14. Note that E10 shares the principal axes of E20 . Hence E 0 (i.e.,
the mapped covered ellipsoid E) has the same principal directions as E20 ; and

36
 (a) (b)

E2 E1 E10

E E0
E20

Figure 14: (a) Sketch of the concentric ellipsoids E1 and E2 and the maximal-
volume ellipsoid E which is covered by them (b) Corresponding ellipsoids in the
transformed space in which E10 is the unit ball.

the lengths of its principal axes are the lesser of those of E10 (all of which are
unity) and those of E20 . Thus if
L02 = UΣVT , (154)
is the SVD of L02 , so that
L02 L0T 2 T
2 = UΣ U , (155)
then the Cholesky matrix of E 0 is given by
2
L0 L0T = UΣ̃ UT , (156)
where the singular values Σ̃ii = σ̃i are given by
σ̃i = max(σi , 1), (157)
where σi ≡ Σii .
The Cholesky matrix of E is obtained by inverting the transformation:
L = L1 L0 . (158)
The same algorithm can be used to determine the minimal-volume el-
lipsoid which covers E1 and E2 . In that case, in contrast to Eq.(157), the
appropriate singular values are:
σ̄i = min(σi , 1). (159)

37
18 Ellipsoid that covers two given ellipsoids
Given two ellipsoids E1 and E2 , the task is to determine a third ellipsoid E
that covers both E1 and E2 . Ideally E is of minimal volume.
It is a problem of convex optimization to determine the minimum-volume
covering ellipsoid (see, e.g., Boyd and Vandenberghe 2004). An algorithm is
provided by Yildirim (2006). It appears that the solution to this convex
optimization problem is computationally expensive. Instead, in the subsec-
tions, we describe heuristic algorithms with determine ellipsoids E (not of
minimal volume) which cover E1 and E2 . These methods are implemented
in the routine ell pair cover, with the parameter algorithm determining the
particular algorithm to be used.

18.1 Spheroid algorithm

The ellipsoids E1 and E2 have centers c1 and c2 , and outer radii rout,1 and
rout,2 . Thus the ball B1 centered at c1 of radius rout,1 covers E1 , and similarly
we define the ball B2 which covers E2 . In this “spheroid algorithm” (algorithm
= 1) we take E to be the minimum-volume ball B which covers B1 and B2 .
The center and radius of B are determined as follows. Consider the line
of centers with distance s measured from c1 towards c2 . Let the distance
between the centers be ∆c ≡| c2 − c1 |. There are four intersections between
the ball B1 and B2 and the line. The outermost of these corresponds to

smax = max(rout,1 , ∆c + rout,2 ), (160)

and

smin = min(−rout,1 , ∆c − rout,2 ). (161)

Thus the center of B is
1
c = c1 + (smin + smax )(c2 − c1 )/∆c, (162)
2
and its radius is
1
r = (smax − smin ). (163)
2

38
 A variant is the “spheroid algorithm with shrinking” (algorithm = 4), in
which r is decreased to r0 to yield the ball B 0 of minimum volume centered at
c which covers E1 and E2 . The radius r0 is determined as the greatest distance
from c to any point in E1 and E2 (which is determined by ell pt near far).

18.2 Covariance algorithm

The ellipsoid E1 is defined by

E1 ≡ {x | (x − c1 )T A1 (x − c1 ) ≤ 1}, (164)

and E2 and E are similarly defined by c2 and A2 , and by c and A, respec-

tively.
In the “covariance” algorithm described in this section, we define the
covering ellipsoid E by
c ≡ 21 (c1 + c2 ), (165)

A ≡ αA0 , (166)
and ³ ´−1
A0 ≡ A−1 −1 1
1 + A2 + 4 [c1 − c2 ][c1 − c2 ]
T
, (167)
where α is a positive parameter to be determined.
To determine α, we consider the ellipsoid E0 defined by c and A0 . With
L0 being the Cholesky factor

L0 LT0 = A0 , (168)

we perform the linear transformation

y = c + L−T
0 x. (169)

As depicted in Fig. 15, this transforms E0 to the unit ball, and E1 and E2
to ellipsoids denoted by E10 and E20 .
Using the furthest-point algorithm (see Sec. 9), we determine y1 and y2 ,
defined as the points on E10 and E20 , respectively, which are furthest from the
origin (see Fig. 15). Clearly, the ball E 0 of radius

r ≡ max(ky1 k, ky2 k) (170)

covers E10 and E20 . This corresponds to E 0 (the transformation of E) with

39
 E0
E10

E00

y1 y2
E20

Figure 15: In the transformed space, the ellipsoid E00 which is the unit ball, the
ellipsoids E10 and E20 , and the covering ellipsoid E 0 .

α = r−2 . (171)

In summary the ellipsoid E which covers E1 and E2 is defined by c

Eq.(165) and A Eq.(166), where A0 and α are given by Eqs. (167), (170), and
(171). The routine ell pair cover includes two implementations of this covari-
ance algorithm. The first (algorithm = 5) is a direct implementation of the
above equations in which A0 is formed and then L0 is obtained by Cholesky
decomposition. The second implementation (algorithm = 2), now described,
is preferred, since it avoids the formation of A0 , and hence is significantly
more accurate.
With L1 and L2 being the Cholesky factors of A1 and A2 , and with
d ≡ 12 (c1 − c2 ), we define the (2n + 1) × n matrix B by

BT = [L−T
1 L−T
2 d], (172)

so that the inverse of Eq.(167) can be written

A−1 −T −1 T
0 = L0 L0 = B B. (173)

Now let the QL factorization of B be

" #
0
B=Q , (174)
L

so that

40
 BT B = LT L. (175)
By comparing the above two equations, we see that L0 is the inverse of
L. Thus, in this preferred QL implementation of the covariance algorithm,
the Cholesky matrix L0 required in Eq.(169) is obtained as L−1 , where L is
obtained from the QL factorization of B, defined by Eq.(172).

18.3 Iterative algorithm

The algorithm described here (which is also implemented in the routine
ell pair cover with algorithm= 2), is more elaborate and expensive than the
“covariance” algorithm described in the previous subsection, but in most
circumstances it generates a covering ellipsoid E of smaller volume.
Given E1 and E2 , the algorithm proceeds through six stages (described
below) to generate the covering ellipsoid E. Stages 1, 2 and 6 are trivial,
but are retained for consistency with the implementation in ell pair cover. In
Stages 3 and 4 the shape of E (but not its center and size) are determined.
The center of E is taken to be on the line of centers of E1 and E2 . Its location
is determined iteratively in Stage 5 so as to minimize the volume of E.
Various spaces are considered, and are referred to as x-space, y-space,
z-space and ζ-space. The ellipsoids E1 and E2 are given in x-space, and the
covering ellipsoid E is to be determined in this space. The other spaces are
obtained by successive linear transformations; and E2 (z), for example, de-
notes the ellipsoid E2 viewed in z-space, which has center c2 (z) and Cholesky
triangle L2 (z).

18.3.1 Stage 1
The two ellipsoids E1 and E2 are given in x-space (in terms of c1 , L1 , c2 and
L2 ): see Fig.16(a).

18.3.2 Stage 2
We denote by E10 and E20 the two given ellipsoids shifted to the origin: see
0
Fig.16(b). (In general, Em denotes Em shifted to the origin.)

18.3.3 Stage 3
The transformed variable y is defined by

41
 (a) E1 (b) E10

O
c1
E2
E20
c2

Figure 16: In x-space, sketches of (a) the ellipsoids E1 and E2 in stage 1 (b) the
ellipsoids E10 and E20 in stage 2.

E3

E1 E2
O
E20
c2 (y)

Figure 17: In y-space, sketches of the various ellipsoids in stage 3.

y ≡ L−1
1 (x − c1 ), (176)
so that E1 (y) is the unit ball at the origin.
At this stage a test is performed to determine if E1 covers E2 ; if it does,
then E = E1 is the minimal volume covering ellipsoid. Similarly, a test is
performed to determine if E2 covers E1 . The way in which this testing is
performed is described in Section 18.3.7. If neither E1 nor E2 covers the
other, then the algorithm proceeds.
The ellipsoid E3 is defined to be the minimal-volume ellipsoid which cov-
ers E10 and E20 (see Fig.17). This is readily determined: E3 (y) has the same
principal directions as E20 , and the lengths of its principal semi-axes are the
greater of those of E20 (y) and E10 (y) (which are unity).

18.3.4 Stage 4
A transformation (to z-space) is performed which consists of a rotation such
that the principal axes of E1 , E2 and E3 are aligned with the coordinate

42
 smax
E2

B2

E1 E4

smin
E3 = B1
L

Figure 18: In z-space, sketch of ellipsoids in stage 4 showing the construction of

E4 based on the extreme intersections (smin and smax ) of the bounding balls B1
and B2 with the line of centers L.

axes, followed by a stretching in the coordinate directions to make E3 (z) the

unit ball at the origin (see Fig.18). This transformation is readily determined
from the SVD of L2 (y). Note that the principal semi-axes of E1 (z) and E2 (z)
are aligned with the coordinate directions; their lengths are at most unity;
and in each direction the length of the principal semi-axis of E1 (z) and/or
E2 (z) equals unity. The line of centers L between E1 (z) and E2 (z) can be
written
L ≡ {z | z = ws},
where w = c2 (z)/|c2 (z)| is the unit vector from the origin to the center of
E2 (z), and s is the distance along the line.
Let B1 denote the bounding ball of E1 (z). This is centered at the origin
and has radius less than or equal to unity. There are two intersections of B1
and L which occur at distances s1− and s1+ along L. Similarly the bounding
ball B2 of E2 (z) intersects L at distances s2− and s2+ . We define the extrema

43
of these intersections by
smin ≡ min(s1− , s1+ , s2− , s2+ ), (177)
and
smax ≡ max(s1− , s1+ , s2− , s2+ ). (178)
The ellipsoid E4 is now defined to be centered at
s0 ≡ 12 (smin + smax ), (179)
to have a principal semi-axis of length 12 (smax − smin ) in the direction w, and
to have all other principal axes unity. In other words, E4 (z) is formed from
the unit ball at s0 by stretching it in the ±w directions so that it intersects
the extrema smin and smax .

18.3.5 Stage 5
A transformation is performed to ζ-space such that E4 (ζ) is the unit ball at
the origin, and the line of centers L(ζ) is in the first coordinate direction:
L(ζ) = {ζ | ζi = tδi1 }, (180)
where t measures the distance along the line (see Fig.19).
The covering ellipsoid being constructed E(ζ) is a ball of radius r0 cen-
tered at a distance t0 along L. It remains to determine r0 and t0 .
Consider a ball centered at a distance t along L. Let r1 (t) denote the
distance from the ball’s center to the furthest point in E1 (ζ). This can
be determined by the algorithm described in Section 9. Similarly, let r2 (t)
denote the distance to the furthest point in E2 (ζ); and we define
r(t) ≡ max(r1 (t), r2 (t)). (181)
Thus the ball centered at a distance t along L and of radius r(t) covers both
E1 (ζ), and E2 (ζ).
In the definition of E(ζ), we take t0 to be (an approximation to) the value
of t at which r(t) is minimum, and then define r0 ≡ r(t0 )
As one moves along the line L from the center of E1 (ζ) to the center
of E2 (ζ), r1 (t) continually increases and r2 (t) continually decreases. Except
in unusual circumstances, the minimum of r(t) occurs between the centers,
where r1 (t) equals r2 (t). A simple iterative procedure usually determines the
location of the minimum (to reasonable accuracy) in two or three iterations.

44
 r1 (t) r2 (t) E
2

O t L
E1

E4

Figure 19: In ζ-space, sketch showing ellipsoids in stage 5, the general point t on
the line of centers L, and the distances r1 (t) and r1 (t) to the furthest points in E1
and E2 , respectively.

18.3.6 Stage 6
The covering ellipsoid E(ζ) obtained in Stage 5 is transformed to the original
coordinate system to yield the required ellipsoid E(x) which covers E1 and
E2 .

18.3.7 Mutual covering

In Stage 3 it is required to determine if E1 (y) covers E2 (y) and vice versa.
The procedure used is now described.
Let ρ be the distance from the origin to the furthest point in E2 (y). Since
E1 (y) is the unit ball at the origin, E1 covers E2 if, and only if, ρ is less than
or equal to unity.
While ρ can be evaluated using the algorithm described in Section 9, this
involves quadratic minimization and hence is somewhat expensive. As now
described, there are easily computed bounds on ρ so that its exact evaluation
can often be avoided.
Let d be the distance between the centers of E1 (y) and E2 (y), and let
r2,min and r2,max be the lengths of the smallest and largest principal semi-axes
of E2 (y). From the triangle inequality we then have

45
 ρ ≤ d + r2,max , (182)
and
ρ ≥ d + r2,min . (183)
We also have
ρ ≥ r2,max . (184)
Thus, if d + r2,max ≤ 1, then ρ ≤ 1, and so E1 covers E2 . On the other hand,
if d + r2,min > 1 or r2,max > 1, then ρ > 1, and so E1 does not cover E2 . In
the remaining cases, E1 may cover E2 , and so ρ is evaluated.

18.3.8 Discussion
While this algorithm is more elaborate and expensive than that given in
Section 18.2, it generally yields a covering ellipsoid E of smaller volume. In
particular it yields the ellipsoid of minimal volume if E1 and E2 are concentric
or if one covers the other.
The principal computational expenses are:
1. The SVD of L2 (y) performed in Stage 3
2. The quadratic minimization involved in the furthest-point algorithm
(Section 9) which is invoked 0, 1 or 2 times (in Stage 3) to determine
if E1 and E2 cover each other
3. The quadratic minimization involved in the furthest-point algorithm
which is invoked twice per iteration (to evaluate r1 (t) and r2 (t)) in
Stage 5.

19 Conclusions
Algorithms have been described for performing some basic geometric opera-
tions on ellipsoids. A Fortran implementation of these algorithms is provided
by the Ell LIB library.

20 Acknowledgments
I am particularly grateful to Professor Charles Van Loan for his help on
numerous occasion with issues of numerical linear algebra.

46
References
Averick, B. M., R. Carter, and J. More (1993). MINPACK-2. http://www-
fp.mcs.anl.gov/OTC/minpack/summary.html.
Boyd, S. and L. Vandenberghe (2004). Convex Optimization. Cambridge:
Cambridge University Press.
Golub, G. H. and C. F. Van Loan (1996). Matrix Computations (3rd ed.).
Baltimore: Johns Hopkins University Press.
Pope, S. B. (1997). Computationally efficient implementation of combus-
tion chemistry using in situ adaptive tabulation. Combust. Theory
Modelling 1, 41–63.
Yildirim, E. A. (2006). On the minimum volume covering ellipsoid of el-
lipsoids. SIAM Journal on Optimization 17, 621–641.

47