diff f6bf760..6fbd3ac daugavet_l1_exact_moduli_v2.txt
diff --git a/daugavet_l1_exact_moduli_v2.txt b/daugavet_l1_exact_moduli_v2.txt
index 83ea662..9de97d2 100644
--- a/daugavet_l1_exact_moduli_v2.txt
+++ b/daugavet_l1_exact_moduli_v2.txt
@@ -73,7 +73,7 @@ even convex functions of r. Therefore the average and the maximum
are also even convex. Write m_N(t) for the modulus obtained by
replacing the average in H_N by the maximum and then taking the infimum
over unit centers. Positivity of hat-delta_N is AMUC, positivity of
-bar-delta_N is AUC. The elementary inequality (a+b)/2 <= max{a,b}
+bar-delta_N is AUC. The elementary ineuality (a+b)/2 <= max{a,b}
gives hat-delta_N(t)<=m_N(t). For the reverse comparison, given a
finite-codimensional subspace F and a unit vector x, intersect F
with the kernel of a norm-one functional phi with phi(x)=1. Each
@@ -226,7 +226,7 @@ defining functional is zero, H=B_X and y=w works. Otherwise divide
l and alpha by ||l|| so that ||l||=1 and alpha<1. Fix
0<eta<min{1-alpha, rho/2}. The rank-one operator T y = l(y) w has
norm one, so there is y in S_X with ||y+l(y) w||>2-eta. Replacing
-y by -y leaves this inequality and the norm of y unchanged, and
+y by -y leaves this ineuality and the norm of y unchanged, and
makes l(y)>=0. Then l(y)>1-eta, because otherwise
||y+l(y)w||<=||y||+l(y)<=2-eta. Hence l(y)>alpha, and
@@ -243,7 +243,7 @@ norming w+y takes value at least 1-eta on each summand. Therefore
>= (1-eta) + a (1-eta)^2.
Letting eta decrease to zero gives ||Id+T||>=1+||T||. The identity
-||Id+T||<=1+||T|| is the triangle inequality, and the case T=0 is
+||Id+T||<=1+||T|| is the triangle ineuality, and the case T=0 is
immediate.
For the passage from slices to weak neighborhoods, the following
@@ -358,7 +358,7 @@ Since t - lambda L_eps = (t-L_eps)_+,
x + t y_eps = (1-lambda) x + lambda (x+z_eps) + (t-lambda L_eps) y_eps,
-and the triangle inequality gives
+and the triangle ineuality gives
||x+t y_eps|| <= (1-lambda) + lambda (1+eps) + (t-L_eps)_+
<= 1 + eps + (t-L_eps)_+.
@@ -379,7 +379,7 @@ satisfies phi(x+ty)=1 and ||x+ty||>=t-1, hence
||x+ty||-1>=max{0,t-2}. So D_E(x,t)>=max{0,t-2}. Together with the
clipping lemma, both lower bounds H_E(x,t)>=max{t/2,t-1} and
D_E(x,t)>=max{0,t-2} hold at every unit center, and the four
-inequalities assemble into the two equalities of the main theorem.
+ineualities assemble into the two equalities of the main theorem.
Failure of AUC renorming. The averaged modulus grows linearly near zero,
whereas the one-sided modulus stays zero up to t=2. So the inherited
@@ -398,7 +398,7 @@ bar-delta_N(t_0) at x_0/N(x_0) provides F in cof(E) with
N(x_0+z) > (1+gamma) N(x_0) when z in F and N(z)=t_0 N(x_0).
-To pass from equality to inequality, use the convexity of
+To pass from equality to ineuality, use the convexity of
r -> N(x_0/N(x_0)+r w) for N(w)=1: for every r>=t_0,
N(x_0/N(x_0)+r w) - 1
@@ -446,9 +446,9 @@ Lebesgue measure P. The metric
d(f,g) = inf { a>0 : P(|f-g|>a)<a }
induces convergence in measure on measurable functions modulo equality
-almost everywhere. The union bound gives the triangle inequality:
+almost everywhere. The union bound gives the triangle ineuality:
P(|f-h|>a+b) <= P(|f-g|>a)+P(|g-h|>b) <= a+b. The identity d(f,g)=0
-holds exactly when f=g almost everywhere. Markov's inequality gives
+holds exactly when f=g almost everywhere. Markov's ineuality gives
d(f,g)<=||f-g||^{1/2}, so norm convergence implies convergence for d.
A subspace depends on finitely many coordinates if one finite coordinate
set determines all its elements. Write B(G) and S(G) for the closed
@@ -592,7 +592,7 @@ Independence of D from the old coordinates gives
so ||psi||<=mr+(1-mq)^{-1}. Put psi_0=psi/max{1,||psi||} in B(G).
The cost of normalization is ||psi-psi_0||<=mr+(1-mq)^{-1}-1. Markov's
-inequality restricted to D and the bound on its complement give
+ineuality restricted to D and the bound on its complement give
P(|g-psi_0|>h) <= mq + h^{-1}( mr + ||psi-psi_0|| )
<= mq + h^{-1}( 2mr + (1-mq)^{-1} - 1 ) < h.