Uniformly and strongly convex functions #
In this file, we define uniformly convex functions and strongly convex functions.
For a real normed space E
, a uniformly convex function with modulus φ : ℝ → ℝ
is a function
f : E → ℝ
such that f (t • x + (1 - t) • y) ≤ t • f x + (1 - t) • f y - t * (1 - t) * φ ‖x - y‖
for all t ∈ [0, 1]
.
A m
-strongly convex function is a uniformly convex function with modulus fun r ↦ m / 2 * r ^ 2
.
If E
is an inner product space, this is equivalent to x ↦ f x - m / 2 * ‖x‖ ^ 2
being convex.
TODO #
Prove derivative properties of strongly convex functions.
A function f
from a real normed space is uniformly convex with modulus φ
if
f (t • x + (1 - t) • y) ≤ t • f x + (1 - t) • f y - t * (1 - t) * φ ‖x - y‖
for all t ∈ [0, 1]
.
φ
is usually taken to be a monotone function such that φ r = 0 ↔ r = 0
.
Equations
- One or more equations did not get rendered due to their size.
Instances For
A function f
from a real normed space is uniformly concave with modulus φ
if
t • f x + (1 - t) • f y + t * (1 - t) * φ ‖x - y‖ ≤ f (t • x + (1 - t) • y)
for all t ∈ [0, 1]
.
φ
is usually taken to be a monotone function such that φ r = 0 ↔ r = 0
.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Alias of the reverse direction of uniformConvexOn_zero
.
Alias of the reverse direction of uniformConcaveOn_zero
.
A function f
from a real normed space is m
-strongly convex if it is uniformly convex with
modulus φ(r) = m / 2 * r ^ 2
.
In an inner product space, this is equivalent to x ↦ f x - m / 2 * ‖x‖ ^ 2
being convex.
Equations
- StrongConvexOn s m = UniformConvexOn s fun (r : ℝ) => m / 2 * r ^ 2
Instances For
A function f
from a real normed space is m
-strongly concave if is strongly concave with
modulus φ(r) = m / 2 * r ^ 2
.
In an inner product space, this is equivalent to x ↦ f x + m / 2 * ‖x‖ ^ 2
being concave.
Equations
- StrongConcaveOn s m = UniformConcaveOn s fun (r : ℝ) => m / 2 * r ^ 2