Optimality Conditions in Convex Optimization A Finite Dimensional View 1st Edition Joydeep Dutta Download
Optimality Conditions in Convex Optimization A Finite Dimensional View 1st Edition Joydeep Dutta Download
https://ebookgate.com/product/optimality-conditions-in-convex-
optimization-a-finite-dimensional-view-1st-edition-joydeep-dutta/
https://ebookgate.com/product/convex-optimization-1st-edition-stephen-
boyd/
ebookgate.com
https://ebookgate.com/product/convex-optimization-in-signal-
processing-and-communications-1st-edition-daniel-p-palomar/
ebookgate.com
https://ebookgate.com/product/finite-dimensional-algebras-and-quantum-
groups-bangming-deng/
ebookgate.com
https://ebookgate.com/product/semidefinite-optimization-and-convex-
algebraic-geometry-content-added-1st-edition-grigoriy-blekherman/
ebookgate.com
The Finite Element Method for Three Dimensional
Thermomechanical Applications 1st Edition Guido Dhondt
https://ebookgate.com/product/the-finite-element-method-for-three-
dimensional-thermomechanical-applications-1st-edition-guido-dhondt/
ebookgate.com
https://ebookgate.com/product/heterogeneous-networks-in-lte-
advanced-1st-edition-joydeep-acharya/
ebookgate.com
https://ebookgate.com/product/adaptationism-and-optimality-1st-
edition-steven-hecht-orzack/
ebookgate.com
https://ebookgate.com/product/a-first-course-in-optimization-1st-
edition-charles-l-byrne/
ebookgate.com
OPTIMALITY CONDITIONS
IN
CONVEX OPTIMIZATION
A Finite-Dimensional View
Anulekha Dhara
Joydeep Dutta
This book contains information obtained from authentic and highly regarded sources. Reasonable
efforts have been made to publish reliable data and information, but the author and publisher cannot
assume responsibility for the validity of all materials or the consequences of their use. The authors and
publishers have attempted to trace the copyright holders of all material reproduced in this publication
and apologize to copyright holders if permission to publish in this form has not been obtained. If any
copyright material has not been acknowledged please write and let us know so we may rectify in any
future reprint.
Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced,
transmitted, or utilized in any form by any electronic, mechanical, or other means, now known or
hereafter invented, including photocopying, microfilming, and recording, or in any information stor-
age or retrieval system, without written permission from the publishers.
For permission to photocopy or use material electronically from this work, please access www.copy-
right.com (http://www.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC), 222
Rosewood Drive, Danvers, MA 01923, 978-750-8400. CCC is a not-for-profit organization that pro-
vides licenses and registration for a variety of users. For organizations that have been granted a pho-
tocopy license by the CCC, a separate system of payment has been arranged.
Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are
used only for identification and explanation without intent to infringe.
Visit the Taylor & Francis Web site at
http://www.taylorandfrancis.com
and the CRC Press Web site at
http://www.crcpress.com
List of Figures xi
Foreword xv
Preface xvii
vii
Bibliography 413
Index 423
xi
xiii
xv
Stephan Dempe
Freiberg, Germany
xvii
Anulekha Dhara
Avignon, France
Joydeep Dutta
Kanpur, India
1.1 Introduction
Optimization is the heart of applied mathematics. Various problems encoun-
tered in the areas of engineering, sciences, management science, and economics
are based on the fundamental idea of mathematical formulation. Optimiza-
tion is an essential tool for the formulation of many such problems expressed
in the form of minimization of a function under certain constraints like in-
equalities, equalities, and/or abstract constraints. It is thus rightly considered
a science of selecting the best of the many possible decisions in a complex
real-life environment.
Even though optimization problems have existed since very early times,
the optimization theory has settled as a solid and autonomous field only in
recent decades. The origin of analytic optimization lies in the classical calculus
of variations and is interrelated with the development of calculus. The very
concept of derivative introduced by Fermat in the mid-seventeenth century via
the tangent slope to the graph of a function was motivated by solving an op-
timization problem, leading to the Fermat stationary principle. Around 1684,
Leibniz developed a method to distinguish between minima and maxima via
second-order derivatives. The calculus of variations was introduced by Euler
while solving the Brachistochrone problem, which was posed by Bernoulli in
1696. The problem is stated as “Given two points x and y in the vertical plane.
A particle is allowed to move under its own gravity from x to y. What should
be the curve along which the particle should move so as to reach y from x in
the shortest time?” In 1759, Lagrange gave a completely different approach
to solve the problems in calculus of variations, today known as the Lagrange
multiplier rule. The Lagrange multipliers are viewed as the auxiliary variables
that are primarily used to derive the optimality conditions for constrained
optimization problems. These optimality conditions are the building blocks of
optimization theory.
During the second world war, Dantzig developed the simplex method to
solve linear programming problems. The first attempt to develop the La-
grange multiplier rules for nonlinear optimization problem was made by Fritz
John [71] in 1948. In 1951, Kuhn and Tucker [73] gave the Lagrange multiplier
rule for convex and other nonlinear optimization problems involving differen-
tiable functions. It was later found that Karush in 1939 had independently
established the optimality conditions similar to those of Kuhn and Tucker.
These optimality conditions are today famous as the Karush–Kuhn–Tucker
(KKT) optimality conditions. All the initial theories were developed with the
differentiability assumptions of the functions involved.
Meanwhile, efforts were made to shed the differentiability hypothesis,
thereby leading to the development of nonsmooth convex analysis as a subject
in itself. This added a new chapter to optimization theory. The key contrib-
utors in the development of convexity theory are Fenchel [45], Moreau [88],
and Rockafellar [97]. An important milestone in this direction was the publi-
cation of Convex Analysis by Rockafellar [97], where the theory of nonsmooth
convex analysis was presented in detail for the first time. No wonder this text
is by far a must for all optimization researchers. In the early 1970s, his stu-
dent Clarke coined the term nonsmooth optimization to categorize the theory
involving nondifferentiable optimization problems. He extended the calculus
rules and applied them to optimization problems involving locally Lipschitz
functions. This was just the beginning. The subsequent decade witnessed a
large development in the field of nonsmooth nonconvex optimization. For de-
tails on nonsmooth analysis, one may refer to Borwein and Lewis [17]; Bor-
wein and Zhu [18]; Clarke [27]; Clarke, Ledyaev, Stern and Wolenshi [28];
Mordukhovich [86]; and Rockafellar and Wets [101].
However, such developments have not overshadowed the importance of
convex optimization, which still is and will remain a pivotal area of research. It
has paved a path not only for theoretical improvements, but also algorithmic
designing aspects. In this book we focus mainly on convex analysis and its
application to the development of convex optimization theory.
C = {x ∈ Rn : gi (x) ≤ 0, i = 1, 2, . . . , m and
hj (x) = 0, j = 1, 2, . . . , l},
φ(x) = ha, xi + b,
where a ∈ Rn and b ∈ R.
It is important to note at the very outset that in optimization theory it is
worthwhile to consider extended-valued functions, that is, functions that take
values in R̄ = R∪{−∞, +∞}. The need to do so arises when we seek to convert
a constrained optimization problem into an unconstrained one. Consider for
example the problem (CP ), which can be restated as
where
f (x), x ∈ C,
f0 (x) =
+∞, otherwise.
All the modern books on convex analysis beginning with the classic Convex
Analysis by Rockafellar [97] follow this framework. However, when we include
infinities, we need to know how to deal with them. Most rules with infinity
are intuitively clear except possibly 0 × (+∞) and ∞ − ∞. Because we will
be dealing mainly with minimization problems, we will follow the convention
0 × (+∞) = (+∞) × 0 = 0 and ∞ − ∞ = ∞. This convention was adopted in
Rockafellar and Wets [101] and we shall follow it. However, we would like to
ascertain that we really need not get worried about ∞ − ∞ as the functions
considered in this book are real-valued or proper functions. An extended-
valued function φ : Rn → R̄ is said to be a proper function if φ(x) > −∞ for
every x ∈ Rn and dom φ is nonempty where dom φ = {x ∈ Rn : φ(x) < +∞}
is the domain of φ.
It is worthwhile to note that the definition of a convex function given
above can be extended to the case when φ is an extended-valued function. An
extended-valued function φ : Rn → R̄ is a convex function if for any x, y ∈ Rn
and λ ∈ [0, 1],
with the convention that ∞ − ∞ = +∞. A better way to handle the convexity
of an extended-valued convex function is to use its associated geometry. In
this direction we describe the epigraph of a function φ : Rn → R̄, which is
given as
F1 + F2 = {x1 + x2 ∈ Rn : x1 ∈ F1 , x2 ∈ F2 }.
λF = {λx ∈ Rn : x ∈ F }.
Bδ (x̄) = x̄ + δB.
The above inequality holds as equality if and only if x = αy for some scalar
α ∈ R.
converges to the i-th component of x̄. The vector x̄ is called the limit of {xk }.
Symbolically it is expressed as
xk → x̄ or lim xk = x̄.
k→∞
It is obvious that the sequences {zr } and {yr } are nondecreasing and non-
increasing, respectively. If {xk } is bounded below or bounded above, the se-
quences {zr } or {yr }, respectively, have a limit. The limit of {zr } is called
the limit infimum or lower limit of {xk } and denoted by lim inf k→∞ xk , while
that of {yr } is called the limit supremum or upper limit of {xk } and denoted
by lim supk→∞ xk . Equivalently,
For a sequence {xk }, lim inf k→∞ xk = −∞ if the sequence is unbounded below
while lim supk→∞ xk = +∞ if the sequence is unbounded above. Therefore,
{xk } converges to x̄ if and only if
Equivalently,
where the term on the right-hand side of the inequality denotes the limit
infimum or the lower limit of the function φ defined as
lim inf φ(x) = lim inf φ(x).
x→x̄ δ↓0 x∈Bδ (x̄)
Equivalently,
φ(x̄) ≥ lim sup φ(x),
x→x̄
where the term on the right-hand side of the inequality denotes the limit
supremum or the upper limit of the function φ defined as
lim sup φ(x) = lim sup φ(x).
x→x̄ δ↓0 x∈Bδ (x̄)
Alternatively, φ is continuous at x̄ if for any ε > 0 there exists δ(ε, x̄) > 0
such that
|φ(x) − φ(x̄)| ≤ ε whenever kx − x̄k < δ(ε, x̄).
The function φ is continuous over a set F ⊂ Rn if φ is continuous at every
x̄ ∈ F .
The next result from Rockafellar and Wets [101] gives a characterization
of limit infimum of an arbitrary extended-valued function.
Proof. Suppose that lim inf x→x̄ φ(x) = ᾱ. We claim that for xk → x̄ with
φ(xk ) → α, α ≥ ᾱ. As xk → x̄, for any δ > 0, there exists kδ ∈ N such that
xk ∈ Bδ (x̄) for every k ≥ kδ . Therefore,
Because δ is arbitrarily chosen, so taking the limit δ ↓ 0 along with the defin-
ition of the limit infimum of φ leads to
which reduces the preceding condition to φ(x̄) ≤ ᾱ, thereby proving that epi φ
is a closed set in Rn × R.
Next we show that (ii) implies (iii). For a fixed α ∈ R, suppose
that {xk } ⊂ lev≤α φ such that xk → x̄. Therefore, φ(xk ) ≤ α, that is,
(xk , α) ∈ epi φ. By (ii), epi φ is closed, which implies (x̄, α) ∈ epi φ, that
is, φ(x̄) ≤ α. Thus, x̄ ∈ lev≤α φ, thereby yielding condition (iii).
Finally, to obtain the equivalence, we will establish that (iii) implies (i).
To show that φ is lsc, we need to show that for every x̄ ∈ Rn ,
On the contrary, assume that for some x̄ ∈ Rn and some sequence xk → x̄,
Thus, there exists a subsequence, without relabeling, say {xk } such that
φ(xk ) ≤ α for every k ∈ N, which implies xk ∈ lev≤α φ. By (iii), the lower
level set lev≤α φ is closed and hence x̄ ∈ lev≤α φ, that is, φ(x̄) ≤ α, which
contradicts (1.1). Therefore, φ is lsc over Rn .
The proof of the last implication, that is, (iii) implies (i) of Theorem 1.9
by contradiction was from Bertsekas [12]. We present an alternative proof for
the same from Rockafellar and Wets [101].
It is obvious that for any x̄ ∈ Rn ,
Proof. Suppose that (x̄, ᾱ) ∈ cl epi φ, which implies that there exists
{(xk , αk )} ⊂ epi φ such that (xk , αk ) → (x̄, ᾱ). Thus, taking the limit as
k → +∞, the condition
yields
as desired.
Conversely, assume that lim inf x→x̄ φ(x) ≤ ᾱ but (x̄, ᾱ) 6∈ cl epi φ.
We claim that, lim inf x→x̄ φ(x) = ᾱ. On the contrary, suppose that
lim inf x→x̄ φ(x) < ᾱ. As (x̄, ᾱ) ∈
/ cl epi φ, there exists δ̄ > 0 such that for
every δ ∈ (0, δ̄),
which implies for every (x, α) ∈ Bδ ((x̄, ᾱ)), φ(x) > α. In particular for
(x, ᾱ) ∈ Bδ ((x̄, ᾱ)), φ(x) > ᾱ, that is,
Therefore, taking the limit as δ → 0 along with the definition of limit infimum
of a function yields
which is a contradiction. Therefore, lim inf x→x̄ φ(x) = ᾱ. By Lemma 1.8, there
exists a sequence xk → x̄ such that φ(xk ) → ᾱ. Because (xk , φ(xk )) ∈ epi φ,
(x̄, ᾱ) ∈ cl epi φ, thereby reaching a contradiction and hence the result.
Now the question is whether it is possible to construct a function that is
the closure of the epigraph of another function. This leads to the concept of
closure of a function.
Definition 1.11 For any function φ : Rn → R̄, an lsc function that is con-
structed in such a way that its epigraph is the closure of the epigraph of φ is
called the lower semicontinuous hull or the closure of the function φ and is
denoted by cl φ. Therefore,
epi(cl φ) = cl epi φ.
−1 1 −1 1
epi φ epi cl φ
If φ is lsc, then it is closed as well. Also cl φ is lsc and the greatest of all
lsc functions ψ such that ψ(x) ≤ φ(x) for every x ∈ Rn . From Theorem 1.9,
one has that closedness is the same as lower semicontinuity over Rn . In this
discussion, the function φ was defined over Rn . But what if φ is defined over
some subset of Rn . Then one cannot talk about the lower semicontinuity of
the function over Rn . In such a case, how is the closedness of a function related
to lower semicontinuity? This issue was addressed by Bertsekas [12]. Consider
a set F ⊂ Rn and a function φ : F → R̄. Observe that here we define φ over
the set F and not Rn . The function φ can be extended over Rn by defining a
function φ̄ : Rn → R̄ as
φ(x), x ∈ F,
φ̄(x) =
+∞, otherwise.
Note that both the extended-valued functions φ and φ̄ have the same epigraph.
Thus from the above discussion, one has φ is closed if and only if φ̄ is lsc over
Rn . Also observe that the lower semicontinuity of φ over dom φ is not sufficient
for φ to be closed. In addition, one has to assume the closedness of dom φ.
To emphasize this fact, let us consider a simple example. Consider φ : R → R̄
defined as
0, x ∈ (−1, 1),
φ(x) =
+∞, otherwise.
Here, dom φ = (−1, 1) over which the function is lsc but epi φ is not closed
and hence, φ is not closed. The closure of φ is given by
0, x ∈ [−1, 1],
cl φ(x) =
+∞, otherwise.
Observe that in Figure 1.1, epi φ is not closed while epi cl φ is closed. There-
fore, we have the following result from Bertsekas [12].
Observe that for a coercive function, every nonempty lower level set is
bounded. Below we prove the Weierstrass Theorem.
(ii) there exists α ∈ R such that the lower level set lev≤α φ is nonempty and
bounded.
(iii) φ is coercive.
Proof. Suppose that condition (i) holds, that is, dom φ is bounded. Because
φ is proper, φ(x) > −∞ for every x ∈ Rn and dom φ is nonempty. Denote
φinf = inf x∈Rn φ(x), which implies φinf = inf x∈dom φ φ(x). Therefore, there
exists a sequence {xk } ⊂ dom φ such that φ(xk ) → φinf . Because dom φ is
bounded, {xk } is a bounded sequence, which by Bolzano–Weierstrass Theo-
rem, Proposition 1.3, has a convergent subsequence. Without loss of generality,
assume that xk → x̄. By the lower semicontinuity of φ,
o(kx − x̄)k
where limx→x̄ = 0. A function φ is differentiable if it is differen-
kx − x̄k
tiable at every x ∈ Rn . The derivative, ∇φ(x̄), of φ at x̄ is also called the
gradient of φ at x̄, which can be expressed as
∂φ ∂φ ∂φ
∇φ(x̄) = (x̄), (x̄), . . . , (x̄) ,
∂x1 ∂x2 ∂xn
∂φ
where , i = 1, 2, . . . , n denotes the i-th partial derivative of φ. If φ is
∂xi
continuously differentiable, that is, the map x 7→ ∇φ(x) is continuous over
Rn , then φ is called a smooth function. If φ is not smooth, it is called a
nonsmooth function.
Similar to the first-order differentiability, we have the second-order differ-
entiability notion as follows.
which is equivalent to
φ(x) = φ(x̄) + h∇φ(x̄), x − x̄i + h∇2 φ(x̄)(x − x̄), x − x̄i + o(kx − x̄k2 ).
The matrix ∇2 φ(x̄) is also referred to as the Hessian with the ij-th entry of
∂2φ
the matrix being the second-order partial derivative (x̄). If φ is twice
∂xi ∂xj
2
continuously differentiable, then the matrix ∇ φ(x̄) is a symmetric matrix.
In the above definitions we considered the function φ to be a scalar-valued
function. Next we define the notion of differentiability for a vector-valued
function Φ.
∂φi
with the ij-th entry of the matrix being the partial derivative (x̄). In
∂xj
the above expression of JΦ(x̄), the vectors ∇φ1 (x̄), ∇φ2 (x̄), . . . , ∇φm (x̄) are
written as row vectors.
Observe that the derivative is a local concept and it is defined at a point
x if x ∈ int dom φ. Below we state the Mean Value Theorem, which plays a
pivotal role in the study of optimality conditions.
With all these basic concepts we now move on to the study of convexity.
The importance of convexity in optimization stems from the fact that when-
ever we minimize a convex function over a convex set, every local minimum
is a global minimum. Many other issues in optimization depend on convexity.
However, convex functions suffer from the drawback that they need not be
differentiable at every point of their domain of definition and the nondiffer-
entiability may be precisely at the point where the minimum is achieved. For
instance, consider the minimization of the absolute value function, |x|, over R.
At the point of minima, x̄ = 0, the function is nondifferentiable. How this ma-
jor difficulty was overcome by the development of a completely different type
of analysis is possibly one of the most thrilling developments in optimization
theory. This analysis depends on set-valued maps, which we briefly present
below.
Definition 1.19 A set-valued map Φ from Rn to Rm associates every x ∈ Rn
to a set in Rm ; that is, for every x ∈ Rn , Φ(x) ⊂ Rm . Symbolically it is
expressed as Φ : Rn ⇉ Rm . A set-valued map is associated with its graph
defined as
gph Φ = {(x, y) ∈ Rn × Rm : y ∈ Φ(x)}.
Φ is said to be a proper map if there exists x ∈ Rn such that Φ(x) 6= ∅. Φ is said
to be closed-valued or convex-valued or bounded-valued if for every x ∈ Rn , the
sets Φ(x) are closed or convex or bounded, respectively. Φ is locally bounded
at x̄ ∈ Rn if there exists δ > 0 and a bounded set F ⊂ Rn such that
Φ(x) ⊂ V, ∀ x ∈ Bδ (x̄).
The set-valued map Φ is said to be closed if it has a closed graph; that is, for
any sequence {xk } ⊂ Rn with xk → x̄ and yk ∈ Φ(xk ) with yk → ȳ, ȳ ∈ Φ(x̄).
A set-valued map Φ : Rn → Rm is said to be upper semicontinuous (usc) at
x̄ ∈ Rn if for any ε > 0, there exists δ > 0 such that
Φ(x) ⊂ Φ(x̄) + εB, ∀ x ∈ Bδ (x̄),
where the balls are in the respective spaces. If Φ is locally bounded and has a
closed graph, then it is usc. If Φ is single-valued, that is, Φ(x) is singleton for
every x, the upper semicontinuity of Φ coincides with continuity.
For more on set-valued maps, the readers may refer to Berge [10]. A de-
tailed analysis of convex function appears in Chapter 2.
(y, f (y))
x y x
The result is obtained by simply multiplying (1.3) with λ and (1.4) with
(1 − λ) and then adding them up. This description geometrically means that
the tangent plane should always lie below the graph of the function. For a
convex function f : R → R, it looks something like Figure 1.2. This important
characterization of a convex function leads to the following result.
that is,
f (x) ≥ f (x̄), ∀ x ∈ C,
Remark 1.21 Expressing the optimality condition in the form of (1.5) leads
to what is called a variational inequality. Let F : Rn → Rn be a given function
and C be a closed convex set in Rn . Then the variational inequality V I(F, C)
is the problem of finding x̄ ∈ C such that
hF (x̄), x − x̄i ≥ 0, ∀ x ∈ C.
C λ = λ0
x x̄
hF (y) − F (x), y − xi ≥ 0.
However, when f is a convex function, one has the following pleasant property.
For proof, see Rockafellar [97]. However, the reader should try to prove it on
his/her own. We have shown that when (CP ) has a smooth f , one can write
down a necessary and sufficient condition for a point x̄ ∈ C to be a global
minimizer of (CP ). In fact, as already mentioned, the importance of studying
convexity in optimization stems from the following fact. For the problem (CP ),
every local minimizer is a global minimizer irrespective of the fact whether f
is smooth or not. This can be proved in a simple way as follows. If x̄ is a local
minimizer of (CP ), then there exists δ > 0 such that
f (x) ≥ f (x̄), ∀ x ∈ C ∩ Bδ .
Now consider any x ∈ C. Then it is easy to observe from Figure 1.3 that there
exists λ0 ∈ (0, 1) such that for every λ ∈ (0, λ0 ),
λx + (1 − λ)x̄ ∈ C ∩ Bδ .
Hence
θ(x) ≥ 0, ∀ x ∈ C
that is,
h∇f (y) − ∇f (x), y − xi ≥ 2ρky − xk2 . (1.6)
The property of ∇f given by (1.6) is called strong monotonicity with 2ρ as the
modulus of monotonicity. It is in fact interesting to observe that if f : Rn → R
is a differentiable function for which there exists ρ > 0 such that for every
x, y ∈ Rn ,
Now we request the reader to show that f is strongly convex with modulus
ρ > 0. In fact, if f is strongly convex with ρ > 0 one can also show that ∇f
is strongly monotone with ρ > 0. Thus we conclude that f is strongly convex
with modulus of strong convexity ρ > 0 if and only if ∇f is strongly monotone
with modulus of monotonicity ρ > 0.
It is important to note that one cannot guarantee θ to be finite unless C
has some additional conditions, for example, C is compact. Assume that C
is compact and let x̄ be a solution of the problem (CP ), where f is strongly
convex. (Think why a solution should exist.) Now as f is strongly convex, it
is simple enough to see that x̄ is the unique solution of (CP ). Thus from the
definition of θ, for any x ∈ C and y = x̄,
thereby yielding
which leads to
s
θ(x)
kx − x̄k ≤ .
2ρ
This provides an error bound for (CP ), where f is strongly convex and C is
compact. In this derivation if ∇f was strongly monotone with modulus ρ > 0,
then the error bound will have the expression
s
θ(x)
kx − x̄k ≤ .
ρ
and hence the error bound provided by considering that f is strongly monotone
with modulus 2ρ gives a sharper error bound.
Now the question is can we design a merit function for (CP ) that can be
used to develop an error bound even when C is noncompact. Such a merit func-
tion was first developed by Fukushima [48] for general variational inequalities.
In our context, the function given by
α
θ̂α (x) = sup h∇f (x), x − yi − ky − xk2 , α > 0.
y∈C 2
θ̂α (x) ≥ 0, ∀ x ∈ C
and θ̂α (x) = 0 for x ∈ C if and only if x is a solution of (CP ). Observe that
α
θ̂α (x) = − inf h∇f (x), y − xi + ky − xk2 , α > 0.
y∈C 2
For a fixed x, observe that the function
α
φα
x (y) = h∇f (x), y − xi + ky − xk2
2
is a strongly convex function and is coercive (Definition 1.13). Hence φαx attains
a lower bound on C. The point of minimum is unique as φα x is strongly convex.
Hence for each x, the function φα x has a finite minimum value. Thus θ̂ α (x) is
always finite, thereby leading to the following error bound.
as desired.
The reader is urged to show that under the hypothesis of the above theo-
rem, one can prove a more tighter error bound of the form
s
2θ̂α (x)
kx − x̄k ≤ .
4ρ − α
2.1 Introduction
With the basic concepts discussed in the previous chapter, we devote this
chapter to the study of concepts related to the convex analysis. Convex analy-
sis is the branch of mathematics that studies convex objects, namely, convex
sets, convex functions, and convex optimization theory. These concepts will be
used in the subsequent chapters to discuss the details of convex optimization
theory and in the development of the book.
23
(λ1 + λ2 )F ⊂ λ1 F + λ2 F. (2.1)
(λ1 + λ2 )F ⊃ λ1 F + λ2 F,
which along with the inclusion (2.1) yields the desired equality. Observe that
(ii) and (iii) lead to the convexity of (λ1 + λ2 )F = λ1 F + λ2 F .
From Proposition 2.3, it is obvious that intersection of finitely many closed
half spaces is again a convex set. Such sets that can be expressed in this
form are called polyhedral sets. These sets play an important role in linear
programming problems. We will deal with polyhedral sets later in this chapter.
However, unlike the intersection and sum of convex sets, the union as well
as the complement of convex sets need not be convex. For instance, consider
the sets
Observe from Figure 2.2 that both F1 and F2 along with their intersection are
convex sets but neither their complements nor the union of these two sets is
convex.
To overcome such situations where nonconvex sets come into the picture
in convex analysis, one has to convexify the nonconvex sets. This leads to the
notion of convex combination and convex hull.
x = λ1 x1 + λ2 x2 + . . . + λm xm
Pm
with λi ≥ 0, i = 1, 2, . . . , m and i=1 λi = 1.
0000000
1111111 0
10
10
10
10
10
1 0 1
1
0
1 00
10
1
000
111
0
1
0
1
11
000
10
10
10
1 0
1 0
10
10
1
0
10
1
111
000
1111111
000000011
00 0
1
11
00 0
1
01
1 0
10 1
1 0 10
1
010
10
111
00
11
00
111
000 11
00 111
000
0
100
10
1 0
1 0
100
10
1
F1 111
0000
1
0
1
0
1
0
1
F
0
12
0
1 0
10
1
0
10
1
1111
0000
1111
0000
0
1
11
00 11
00 111
0000
1
11111
000000
10
1 0
1 1111
0000
0
1 0
1
1111
0000
111
000 11
00
111
000 01
10 0
1 0
1
111
000 0 1
1 0 1
1111111111
00000000000
1111111
000000011
00 1111111111
0000000000
11111111111
00000000000
1111111
0000000
1111111
0000000 1111111111
0000000000
F1c F2c
1
0
0
10
1
0
1 0
1
1
0 1
0 1
00
10
1
0
10
1
01
10
0
1 0
1
0
1 0
1 0
1
0
10
10
1
0
10
1
0
1
0
1 0
1
0
1 0
1 0
1
0
1 0
1
0
1
01
1 0
1
0
1
0
1 0
1
1
0
0
1 1
0 1
0
1
0
0
10
10
1
0
0
1
0
1
0
1
1
00
1 0
1
0
1 0
1
0
1 0
1 0
1
0
10
10
1
0
10
1
0
1 0
1 0
1
0
1
1 0
1 0
1
1
00
10
1
0
10
10
1
0
1 0
1 0
0
1 1
0 0
1
0
1
01
100
1
0
1 0
1 0
1
0
1
0
1 0
1
0
1 0
1
0
1
0
1
0
1 11
0
0
1 0 0
1
0
1
0
1
1
0 1
0 1
0
1
0
0
1
0
1 0
1 0
1
0
1
0
1 0
1
0 1
0 1
1 0
F1 ∩ F2 F1 ∪ F2
FIGURE 2.2: F1 , F2 , and F1 ∩ F2 are convex while F1c , F2c , and F1 ∪ F2 are
nonconvex.
The next result expresses the concept of convex set in terms of the convex
combination of its elements.
Theorem 2.5 A set F ⊂ Rn is convex if and only if it contains all the convex
combinations of its elements.
x = λ1 x1 + λ2 x2 + . . . + λl xl ,
Had Henry but known it, there was no immediate necessity for his
return to the Iroquois. Indeed, the Coast Guard cutter not only was
unable to sail that day, but she did not cast off her hawsers until
afternoon the day following. Although Henry thus had an
unexpected half day in Boston, he saw no more of the city than he
had seen on the preceding day, for when he awoke rain was pouring
down, a vicious east wind was blowing, and the elements were as
nasty as they well could be.
Even a complete suit of rubber would hardly have kept one dry
very long in the slashing, blustering blasts that came howling
through the Navy Yard. The rain drove in horizontal sheets. It
whipped around corners and under doorways and awnings. It roared
across open decks. It beat against the air-ports. With ever-increasing
force the gusts came tearing in from the sea. Trees bent far over
and groaned before their onslaught. Flags whipped themselves to
ribbons. Halyards beat a very devil’s tattoo against their masts, and
on the Iroquois the cordage fairly shrieked in wild, wailing notes that
made Henry almost shudder.
He was glad enough that the Iroquois had been unable to get
away. Never had the captain’s cabin seemed half so comfortable and
attractive. He told himself that he would be glad enough to spend
the time there, reading some of the interesting books from the
captain’s bookshelf, while waiting for the storm to subside.
Yet the movement of stores went forward without interruption.
Sailors, cased from head to foot in sou’westers, oilskins, and rubber
boots, worked without ceasing in the downpour to finish the
transshipment. Henry could hear them clumping about the deck in
their clumsy footgear. On the pier trucks rattled and banged. Boxes
were wheeled aboard and dumped on the deck. Men swore and
slipped in the wet. Machinery rattled.
While Henry and his host were eating their luncheon, the noises
suddenly ceased. There were a few shouted orders, indistinguishable
in the roar of the storm, then some banging noises as hatches were
closed and battened down, and other openings made fast. Soon all
was quiet. When the luncheon was ended, Henry went up the
companionway and peeped out. Things had been made tight.
Awnings had been removed. Everything was lashed fast. The decks
were bare. The Iroquois was stripped for action.
Henry could hardly believe that Captain Hardwick would leave
port in such a storm. He knew that storm warnings must be showing
all along the New England coast, and perhaps the entire Atlantic
seaboard, but before he could return to the cabin and ask the
captain if he intended to put to sea, he saw the chief electrician
running aft along the slippery deck. Henry threw open the door for
him, and the radio man dodged in out of the blinding rain. He had a
message for the captain.
Henry descended to the cabin with the electrician. Captain
Hardwick took the telegraph blank from Mr. Sharp and studied it a
moment.
“No orders?” he asked.
“Not yet,” said the electrician.
The captain sat down at his desk, drew a red-bordered telegraph
blank from a pigeonhole, and wrote. “Send that,” he directed.
The radio man struggled out into the storm. The captain rose and
touched his call-bell.
“Rollin,” he said, when his attendant appeared, “ask Mr. Farley to
come to me at once.”
The commander picked up the telegram and handed it to Henry.
“You’ll have a chance to see something today, Henry,” he said.
Henry looked at the message in his hands. It was a cry of
distress, an SOS message the radio man had picked out of the air:
“Steamer Capitol City ashore east
coast Cape Cod forty-two north, seventy
west. Pounding badly. Need immediate
assistance.”
The chief engineer entered the cabin. “Start your engines at once,
Mr. Farley. We just caught a distress call. Steamer ashore on the east
coast of Cape Cod. See that everything is ready for a hard run.”
“All right, sir,” and the chief engineer hurried forward.
Soon the ship began to vibrate. The rumbling noise of the
machinery arose. Yet the Iroquois lay quiet in her dock.
“He’s warming her up,” thought Henry. “It’s a good thing, too, for
this is going to be some struggle. I’ll bet the waves are like
mountains.”
The captain rang his call-bell. “Send Lieutenant Hill to me,” he
said.
The lieutenant came briskly into the cabin. The two officers
conferred a moment. Henry stepped over to the wardroom. Not even
the doctor was there. Cards and books lay on the table in disorder,
as they had been dropped at news of the wireless. Everybody was at
his post, preparing. Henry returned to the cabin. The lieutenant was
gone. The commander was pulling on his oilskins.
“Where can I get a slicker?” Henry inquired.
“Ask Rollin. He will get you one.”
The captain pulled on his sou’wester, tied the strings under his
chin, and mounted to the deck. Henry rang for Rollin and stated his
wish. The attendant brought boots, hat, and slicker. Henry pulled on
a sweater, buttoned his coat up tight, and pulled on the waterproofs.
The gust of wind that struck him as he came out of the
companionway sent him reeling back against that structure. He
could hardly catch his breath. The driving sheets of rain blinded him.
He fought his way forward, and entered the radio shack. The chief
electrician was copying down a message. It was an order for the
Iroquois to go to the rescue of the Capitol City. But Henry knew that
it was not needed. Already the Iroquois was prepared to get under
way.
Henry took the message to the bridge and handed it to the
commander. “Tell Sparks to wire that we are leaving Boston to help
the Capitol City,” shouted the captain.
Henry carried the message to Mr. Sharp and watched him send it.
Then he went back to the bridge. The rain beat on him as irresistibly
as ever, but the weather-cloth offered surprising protection from the
wind. A sailor slipped the hawsers over the posts on the pier. Other
sailors drew in the hawsers and stowed them away. The captain
pressed his signal-bell, and the Iroquois began to move astern. She
backed out into the stream and then turned and headed for the sea,
into the teeth of the driving storm.
The beating rain obscured the view. Fog made the shores almost
indistinguishable, for in from the sea, blown on the breath of the icy
blasts, came racing great clouds of murky white vapor that screened
all they touched. The captain looked grim and inscrutable. His jaw
was set hard. He stood by the wheelhouse, conning the ship. At half
speed the Iroquois slowly nosed her way down the channel. Wiped
from view was the beautiful scene that had so delighted Henry a few
short hours before. Nothing could be seen but occasional glimpses of
shore, the tumultuous, muddy water, and the driving curtains of fog.
One by one the captain made the proper turns in the tortuous
channel. As the Iroquois stood farther and farther out toward the
sea, the waters became ever more tumultuous, the winds roared
more fiercely, and the fog shut in ever denser. Fathom by fathom the
ship crept past one after another of the island defenses along the
way, that served as breakwaters to the sea and broke the sweep of
the winds. When at last the little ship turned eastward at George’s
Island, and faced the storm with the last vestige of protection gone,
she trembled and shook in the grasp of the roaring blasts.
A smother of foam was the sea. Waves rose and broke in
incredible confusion. The waters were churned as by a giant hand.
The racing winds whipped the crests from the combers and flung
them forward in sheets of blinding spray. Fog drove onward in
clouds, now completely hiding the sea, now lifting momentarily, to
expose the wild waste of tossing waters. The fury of the storm was
indescribable.
Mountain high indeed seemed the waves. Before the bow of the
Iroquois they rose up, up, up, as high as the men on the bridge,
then rushed savagely at the little boat, seemingly bent on her
destruction. Down they crashed, and the nose of the cutter was
buried in a smother of foaming water. Sometimes the crests swept
completely over the bow, pouring over the forward deck in great
floods that raced aft and went foaming out of the scuppers. Now
Henry saw why the decks had been cleared of all movable objects.
Indeed, as he watched the smashing combers crash over the bow,
he feared that the big guns themselves would be torn from their
foundations on the iron deck-plates and hurled aft against the
wheelhouse. With blanched face he stood on the bridge, desperately
gripping the rail, and peering with fascinated gaze at the snarling,
hungry seas.
Meantime the captain had sent a reassuring message to the
Capitol City, telling her the Iroquois was on her way to assist her.
And when the ship was fairly in the sea, past all chance of harm by
rock or shoal, the captain left the lieutenant in charge of the bridge
and went himself to the chart-room to plan his coming movements.
With him went Henry. He had seen enough of the sea for a time.
Some of the fear that first gripped his soul had gone. He knew that
the Iroquois was safe so long as she held her course, but he wanted
to shut out for a time the sight of those terrifying billows; so he
staggered to the chart-room, and stepped inside, glad of a relief
from the terrible tension that had held him.
The captain was calmly poring over his charts and guide-books.
“Forty-two north, seventy west,” he muttered, sweeping his glance
over an outstretched map. He placed his pencil on the indicated
spot. “She’s on the shoals almost dead east of Truro,” he said to
Henry.
Then he turned to a Coast Guard directory and leafed it over.
“Thank God!” he cried. “There’s a Coast Guard land station near by.
It may be two or three miles distant, but they’ll come with their
outfit. They must have gotten the wireless flash just as we did. It’s
likely they’re on their way now. What a fight that will be—those
miles across the sand with their heavy boats in this wind. They’ll do
it, though. But maybe we can get to the Capitol City before she
breaks up. I wish we dared go faster.”
He stepped out on the bridge and conferred with the lieutenant.
“Do you think that we dare drive her any faster?” he shouted in the
latter’s ear.
The lieutenant shook his head dubiously. “Well, we’ll try it,” said
the commander.
And he stepped into the wheelhouse and signaled for more speed.
The response was terrifying. Under increased power the Iroquois
drove ahead, bit by bit gaining greater momentum. Savagely she
charged into the seas. Wildly the waves leaped to meet her. The
impact grew ever more terrific. Soon a huge mountain of water
came roaring down on the little cutter. Up it rose and up and up,
while the cutter charged to meet it, and as it broke and crashed
forward, the cutter dived completely under it. The shock was terrific.
Tons of water crashed down on the deck, then went racing aft
against the superstructure. The Iroquois seemed to stop in her
tracks. She trembled from stem to stern. She shook and quivered.
The great wave went roaring aft, twisting and tearing at the ship. A
section of the rail carried away, and went pounding astern. The
tumult was terrifying. Henry, in the chart-room, thought the ship was
breaking to pieces. With blanched face and set teeth he clung to the
side of the room, not knowing what to do. He was reassured when
the captain calmly stepped inside the wheelhouse and signaled for
the engineer to lessen speed.
Hour after hour they wallowed through the storm. Unflinching,
undaunted, in complete mastery of ship and crew, the captain stood
on the bridge, with his right-hand man. Mile by mile, league after
league, he fought his way eastward. The rain ceased, but the winds
blew on, lashing the seas to fiercer and greater fury. The sky was
totally obscured. Gradually the fog lifted, but not for long was the
increased vision. Darkness came apace, and into that roaring
darkness the Iroquois drove with all the power she dared to use.
With anxious eyes the captain watched the logbook. At times he
telephoned to the chief engineer. Again and again he stepped to the
wheelhouse and looked at the compass. Anon he consulted with the
lieutenant. He must know where he was, in order to get safely round
the long arm of Cape Cod. Nor did he dare stand too close to shore
in his run along the eastward coast of the Cape, lest the storm put
the Iroquois also on the shoals. Dead reckoning alone would tell him
when to turn, and carefully the commander considered every feature
that might indicate his position.
Darkness had long since shut in, and the Iroquois was wallowing
through a night as black as pitch before the captain altered his
course and headed south along the outer edge of the Cape.
Gradually the course of the vessel was shifted. To Henry the change
was terrifying. No longer was the Iroquois breasting the storm. The
waves took her abeam. From side to side she rolled until Henry’s
heart stood still with fear. Over and over and over she dipped until
he was certain she would turn upside down. Then slowly she righted
and swung in the opposite direction. And once, when she rolled at
an angle of forty degrees, Henry almost gasped aloud. It seemed
like eternity while the ship lay poised almost on her beam’s ends,
apparently uncertain whether to roll on over or come back on her
keel. Then she slowly righted.
Meantime the chief electrician had been in touch with the
stranded ship. From her signals he knew that she was not far away.
They came crackling out of the air sharp and clear. A distant glow
showed that the guardsmen from the land were already at the
scene. At last the Capitol City wired that she could see the lights of
the Iroquois.
“Make a flare,” flashed back the Iroquois’ wireless man.
And presently, almost dead to leeward of the Iroquois, the
darkness was torn by a flash, and a flaming rocket went streaking up
through the night. Other rockets followed. Then a flaring light arose,
and through their glasses the men on the bridge of the Iroquois
could see the stricken ship, lying in a smother of foam on the outer
edge of the breakers. She was too far from shore for the land crew
to shoot a line to her, and no small boat could live in such a sea. If
the crew of the Capitol City were to be saved, the little cutter alone
could save them.
Cautiously the Iroquois was worked in toward the shoals. Then
slowly she was turned, the captain gave the word, and one of the
anchors was let go. Once more the cutter was heading again into
the seas, and now, little by little, her anchor-chain was paid out,
while seamen with hand-lines took soundings, calling up to the
bridge the depth they found. Fathom after fathom the giant anchor-
chain was paid out. Fathom after fathom the Iroquois rode backward
toward the seething breakers.
Fascinated, Henry watched the attempt to get within reach of the
unfortunate steamer. Gradually the Iroquois drew near to the
smother of white water. The sea was shoaling fast and the tide was
running out, but the captain kept on in the hope that he could get
near enough to shoot a line aboard the Capitol City. His shells, and
the shot-line, loosely wound in the faking-box so that it would run
out freely, were ready for instant use. All that was necessary was to
remove the canvas cover from the little gun on the after rail, insert
the charge, and shoot. But the Iroquois never came within shooting
distance. Too rapidly the water shoaled, and at last, reluctantly, the
commander gave the word and the anchor-chain was held taut. The
Iroquois was rolling, head to the sea, at the very edge of the
breakers, but she was too far away from the Capitol City to put a
shot across her.
“Tell them we’ll float a line down to them, and for them to be on
the lookout for it,” the captain ordered the wireless man.
The latter sat down at his key, but a moment later switched off.
The Capitol City’s wireless was failing. He shifted to the blinkers, and
for the first time Henry had opportunity to see the lights on the
yardarm flash and blink. From the Capitol City came answering winks
from aloft.
“They’re looking for our line,” the wireless man informed the
captain.
Meantime empty kegs had been prepared. A light line was made
fast to one of them, and it was thrown into the sea. Rapidly it
floated to leeward, and as fast as the sea carried it shoreward the
line was paid out. The searchlight of the Iroquois was broken out
and its beam kept on the floating keg. Slowly this bore down on the
Capitol City, but it was too far to one side for that ship to get it. The
line was hauled in, and again it was floated toward the helpless ship.
This time a sailor hurled the keg far to one side of the Iroquois, in
the hope that it might now come close enough to the stranded ship
to be caught. But again the attempt failed. The ebb tide, with its
cross current, carried it farther away from the Capitol City than it
had been before.
“Lower the surfboat,” ordered the commander, when he saw the
attempt was useless. Then he called for volunteers. By the dozen the
sailors leaped forward.
“Boatswain Johnson,” said the captain, “I’m going to put you in
charge. Pick your crew.”
The boatswain selected eight sturdy sailors for oarsmen. A life
belt was strapped round each. The boat was lowered to the rail, and
the crew stepped carefully aboard. At a favorable moment the craft
was launched. Quickly she shot away from the side of the Iroquois,
and before another comber broke, she was at a safe distance from
the cutter and heading straight into the breakers.
The ship’s light was trained on her. On she went, now up, now
down, breasting the roaring waves, shooting through the smother of
foam, riding safely where it seemed impossible for a boat to live,
under the skillful guidance of the experienced boatswain. Swiftly she
drew toward the Capitol City, which no longer lay at right angles to
the beach, but had worked a little to one side, making a lee where
the water was calmer. Toward this the boatswain drove the surf
boat. Into it the little craft shot safely, while a sigh of relief went up
from the deck of the Iroquois.
Through powerful glasses, Henry watched breathlessly while the
surfboat drew close to the protected side of the Capitol City. A line
was thrown to the little boat, and a sailor in the bow caught it. Then
the surfboat was drawn close beside the stranded steamer, and a
sailor scrambled down from the rigging and dropped into it.
From seaward a giant comber was rushing toward the tiny craft. If
it caught her, it would crush her against the side of the larger ship as
an eggshell is crushed underfoot. On the Iroquois not a soul
breathed. The onlookers stood tense, waiting the outcome. But the
boatswain had an eye for everything. He, too, saw the great comber
approaching. Back from the steamer’s side drew the little boat, and
the wave passed harmlessly by. Again the surfboat drew up to the
side of the Capitol City, and another man detached himself from the
rigging and dropped into it. But just at that moment a wave, rising
apparently from nowhere, swept over the windward side of the
stranded steamer, poured irresistibly across her deck, and sent its
tons of icy water crashing downward into the little surfboat lying by
the leeward rail. The tiny craft was swept from sight.
“My God!” cried the captain. “They’ve capsized!”
CHAPTER XI
LOST IN THE SEA
Slowly the staunch little cutter steamed ahead. Had it not been
for the broken rudder, the Capitol City could have gone on her way,
unattended, as she was little damaged. But with her rudder injured,
she was in a bad way. Although it was not entirely gone, and it
helped somewhat in guiding the ship, still at times the huge craft
yawed badly. By paying out more cable Captain Hardwick partly
overcame this tendency to yaw. Nevertheless it was difficult enough
to make headway, for the wind had now whipped around almost to
the north, and for a time the two ships had to sail into the teeth of
it. The Capitol City could not use the full power of her engines, but
she kept her propeller turning, and this greatly lessened the burden
on the cutter. Unless some unforeseen difficulty should arise, the two
vessels would have no trouble in reaching their port.
There was much to be done on both ships, however. The captain
of the disabled steamer wanted to get in touch with his agents, and
inform them of the safety of the vessel. His wireless was still out of
commission, and his messages had first to be sent to the Iroquois by
blinkers and then relayed by wireless. So both Henry and his friend
Belford were kept busy for a long time. Then, too, Henry had to get
in touch with the Coast Guard land station where the men from the
Iroquois had found shelter, and send instructions concerning them.
The captain felt that he could operate the Iroquois with his little
force for the short run to Boston and thence to New York. His men
were willing to do double duty. It was necessary for him to get back
to New York at the earliest possible moment. The run to Boston
would require only a few hours. Then he could retrace his steps
immediately. It was simpler to have the seamen go direct to New
York and rejoin the Iroquois there, than to wait for them in Boston.
So Henry got in touch with the land station and communicated the
captain’s orders to that effect.
By the time the two young wireless men had finished their tasks,
they were about worn out. It was far into the night. Belford had
been on duty for many hours. Henry, though not on duty so long,
was almost exhausted by the nervous strain under which he had
been working. Furthermore, he would have to respond, at any
moment, to any call for wireless communication.
“We must get to bed,” he said the moment their tasks were done.
“Won’t you please call the other operator? Then we can arrange our
watches.”
Belford summoned Black, who came into the radio shack,
yawning. “I guess you’ll have to stand watch for a while, Black,” said
Henry. “Things have got a little mixed up in the wireless house, but
we’ll get them straightened out shortly. Suppose you stay until
breakfast. That won’t be long. Then we can rearrange our watches.”
“That isn’t my regular watch,” said the young radio man, sullenly.
“And why should you be telling me what to do, anyway? That’s up to
Mr. Sharp.”
“But Mr. Sharp isn’t here,” replied Henry. “He went in one of the
boats. I am working with you men. We’ve got to do the best we can
under the circumstances. We’ll get things straightened out soon, and
back in their old order.”
“And do you think I am going to take orders from you?” said the
young operator, insolently.
Henry’s face flushed. An angry light leaped to his eyes, and his fist
involuntarily clenched. But he took a grip on himself, and replied
very slowly and deliberately: “It isn’t a question of taking orders,
Black. The captain has asked me to work with you two men, and the
three of us must pull together until we get the ship safe home.
Aren’t you willing to do your share? Belford here has been working
like a Trojan for I don’t know how many hours. He’s all worn out. We
want you to take this trick. You’ll have to watch close, for there’ll
likely be calls for us. Summon me at once if you hear our call.”
“I reckon I can answer a call as well as anybody on this boat,”
growled the scowling radio man.
“See here, Black,” said Henry. “Why do you want to make trouble?
We’re still at sea, and we’ve got some tall traveling to do before we
are back safe in New York. We’ve all got to pull together. I’m not
asking you to do this for me. It’s for the good of the ship.” And then,
noting the sullen look in the lad’s eyes, Henry continued: “There’s
one thing more. I’m acting under orders from the captain. He has
ordered me to take charge of the wireless. My instructions are to let
no one touch the key but myself. All I can do is to pass those orders
along. If you hear a call for the Iroquois, waken me at once. I’ll
answer it.”
Young Black mumbled an unwilling assent, and swore under his
breath. Henry and the other radio man passed out of the shack to
the stateroom. They noticed that the wind had decreased greatly.
Once in the stateroom, Henry turned to his companion. “What do
you suppose ails that fellow, Black?”
“I don’t know. None of us can make him out. He’s been snappy
and surly ever since he came aboard a few months ago. He’s lazy—
too lazy to study and practice and become a really good operator.
But he is quick and reads wireless calls very well. He spends most of
his time smoking cigarettes and reading dime novels. He’s simply
crazy to use the wireless. He thinks he’s a great operator. The chief
electrician would be glad to let him practice under his own
supervision, the way he does with me. But he doesn’t want either of
us to send when he isn’t present. He’s afraid we might burn out
something. I don’t want to knock Black, but I’d be afraid to have him
monkey with any machine I was responsible for myself. He’s the
most careless, reckless fellow I ever saw.”
“But why should he take such an apparent dislike to me?”
demanded Henry. “I’m a complete stranger to him. He hasn’t any
possible reason for disliking me.”
“He thinks he has,” said Belford.
“He does? What is it?”
“Why, Mr. Sharp let you send a message on the wireless outfit
yourself.”
“But Mr. Sharp was right beside me, to make sure I did no harm.”
“I know it, but that doesn’t make any difference with a fellow like
Black. He dislikes you very much. He tried to tell me a long story
about it. And to have you put over him as chief is more than he can
stand. He’s a bad egg, I believe. And I’d watch him closely if I were
you.”
“Of course I’ll watch him, if he’s likely to shirk his duty,” said
Henry. “He might get the ship into trouble.”
“I don’t mean to watch him in that way. Watch him on your own
account. I don’t know that he’d really do anything to anybody. But
he’s always talking about fixing this fellow or that fellow. He might
try to do you some harm.”
Henry smiled. “There’s little danger,” he said. “We’ll be in New
York in a couple of days, and I’ll probably never see him again after
that.”
But though Henry smiled at the idea of the young operator’s doing
him harm, he could not dismiss from his mind so easily the feeling
that had come to him of uneasiness for the boat itself. He got his
nightclothes from the captain’s cabin, undressed, and slipped into
the chief operator’s bunk. But weary though he was, he could not
sleep. He was worried, and worry was almost a new sensation to
him. He could not at first understand it. He would be almost asleep,
then he would wake up with a start, his leg or arm jerking nervously.
It was partly the reaction from the long strain he had been under.
That sort of experience was absolutely new to him, and he could not
forget the horrible things he had witnessed. But gradually he came
to realize that there was something more to his restlessness than
the reaction from nervous strain. Something was preying on his
mind. After he had tossed sleeplessly on his mattress for a long
time, he comprehended what was the matter. He was worried about
the wireless.
This was the first time in his life that Henry had ever been placed
in a situation of great responsibility. The responsibility was very real,
too, for the lad understood that in large measure the safety of the
ship now depended upon him. Much more than his own life was at
stake. The lives of all the men left on the cutter, and the safety of
the ship herself, might at any moment become absolutely dependent
upon him. If no danger threatened aboard the Iroquois, there might
yet be other lives and other ships endangered, and upon how well
his instruments were cared for, and how thoroughly the watch was
kept, would depend the possibility of catching those far-flung cries
for help.
Had Henry been put in charge of the wireless thus suddenly and
unexpectedly, under less harrowing circumstances, he doubtless
would have felt the responsibility far less poignantly. But for hours
past he had been watching the fight against death, and even death
itself. He was nervously keyed up to a fearfully high pitch. His nerves
would not let down. Doubtless it was the sense of responsibility that
kept them taut. Again and again he told himself that it was his
business to go to sleep and rest, so that he might be fit for duty
when his turn came. But his arguments had no effect upon his
nerves. He was unable to sleep.
For an hour or two he tossed in his bunk restlessly. Again and
again he fell asleep, only to wake a few minutes later. The situation
preyed on him. He could not keep his thoughts away from the radio
shack. Suppose all was not right there. Suppose young Black was
taking this time to defy orders and fool with the wireless key.
Suppose when Henry woke up he should find the wireless out of
commission. And suppose the Iroquois should break her rudder, or
crack her shaft. She might, with such a strain on it, in such a sea.
Suppose the captain should go wrong in his reckoning and the cutter
should pile up on the sands, as the Capitol City had done. Suppose—
suppose——A hundred terrifying suppositions came into Henry’s
mind. Finally he could stand them no longer. He rose, drew on his
clothes, and made his way toward the radio shack. Light was
beginning to illumine the eastern sky. The sea still heaved violently,
but the wind had lost most of its force. Peace was returning to the
troubled world.
Then Henry stepped inside of the radio shack, and stood as
though petrified. Young Black lay back in his chair, his eyes tightly
closed, snoring loudly. He was sound asleep. For all that Henry
knew, he might have been asleep for a long time.
“Black,” said Henry sternly, touching the sleeping wireless man on
the shoulder. When the sleeper merely grunted, Henry said, still
more sternly, “Wake up!” And this time he shook the sleeper
vigorously.
Slowly young Black opened his eyes. A savage oath burst from his
lips as he saw who was standing before him and realized his
situation. Then a crafty look came into his eyes. He laughed.
“Fooled you that time, didn’t I?” he chuckled. “You thought I was
asleep.”
“No, you didn’t fool me,” said Henry. “I don’t think you were
asleep. I know it. Put on your headphones and attend to your
business.”
All the ill nature in the other lad’s being leaped to the front. “You
spy!” he snarled, with an oath. “So that’s the sort of chief radio man
we’ve got, is it? A sneaking spy!”
For a moment it was all Henry could do to keep from hitting the
lad. But he took a grip on himself and for a moment made no
answer. Then he said: “I shall report this matter to the captain as
soon as possible. You may go to your bunk. I will finish your watch
myself.”
Slowly the third-class radio man arose and left the shack. At the
door he turned and faced Henry. “You sneaking spy,” he said, with
another oath, “if you tell the captain, I’ll fix you as sure as my name
is Black.”
Welcome to Our Bookstore - The Ultimate Destination for Book Lovers
Are you passionate about books and eager to explore new worlds of
knowledge? At our website, we offer a vast collection of books that
cater to every interest and age group. From classic literature to
specialized publications, self-help books, and children’s stories, we
have it all! Each book is a gateway to new adventures, helping you
expand your knowledge and nourish your soul
Experience Convenient and Enjoyable Book Shopping Our website is more
than just an online bookstore—it’s a bridge connecting readers to the
timeless values of culture and wisdom. With a sleek and user-friendly
interface and a smart search system, you can find your favorite books
quickly and easily. Enjoy special promotions, fast home delivery, and
a seamless shopping experience that saves you time and enhances your
love for reading.
Let us accompany you on the journey of exploring knowledge and
personal growth!
ebookgate.com