Course on Inverse Problems Albert Tarantola Third Lesson: - - PowerPoint PPT Presentation

course on inverse problems
SMART_READER_LITE
LIVE PREVIEW

Course on Inverse Problems Albert Tarantola Third Lesson: - - PowerPoint PPT Presentation

Princeton University Department of Geosciences Course on Inverse Problems Albert Tarantola Third Lesson: Probability (Elementary Notions) Let u and v be two Cartesian parameters (then, volumetric probabilities and probability densities are


slide-1
SLIDE 1

Princeton University

Department of Geosciences

Course on Inverse Problems

Albert Tarantola

Third Lesson: Probability (Elementary Notions)

slide-2
SLIDE 2

Let u and v be two Cartesian parameters (then, volumetric probabilities and probability densities are identical). Given a probability density f (u, v) , one defines the two marginal probability densities fu(u) =

+∞

−∞ dv f (u, v)

fv(v) =

+∞

−∞ du f (u, v)

and the two conditional probability densities fu|v(u|v = v0) = f (u, v0)

+∞

−∞ du f (u, v0)

fv|u(v|u = u0) = f (u0, v)

+∞

−∞ dv f (v, u0)

slide-3
SLIDE 3

0.02 0.04 0.06 0.08 0.1 0.12

  • 10
  • 5

5 10 0.025 0.05 0.075 0.1 0.125 0.15

  • 10
  • 5

5 10 0.02 0.04 0.06 0.08 0.1 0.12 0.14

  • 10
  • 5

5 10 0.02 0.04 0.06 0.08 0.1 0.12 0.14

  • 10
  • 5

5 10

  • 10
  • 5

5 10

  • 10
  • 5

5 10 0.05 0.1 0.15 0.2 0.25

  • 10
  • 5

5 10 0.05 0.1 0.15 0.2 0.25

joint conditional conditional conditional conditional marginal marginal

slide-4
SLIDE 4

Let u and v be two Cartesian parameters (then, volumetric probabilities and probability densities are identical). Given a probability density f (u, v) , one defines the two marginal probability densities fu(u) =

+∞

−∞ dv f (u, v)

fv(v) =

+∞

−∞ du f (u, v)

and the two conditional probability densities fu|v(u|v = v0) = f (u, v0)

+∞

−∞ du f (u, v0)

fv|u(v|u = u0) = f (u0, v)

+∞

−∞ dv f (v, u0)

slide-5
SLIDE 5

Let u and v be two Cartesian parameters (then, volumetric probabilities and probability densities are identical). Given a probability density f (u, v) , one defines the two marginal probability densities fu(u) =

+∞

−∞ dv f (u, v)

fv(v) =

+∞

−∞ du f (u, v)

and the two conditional probability densities fu|v(u|v0) = f (u, v0)

+∞

−∞ du f (u, v0)

fv|u(v|u0) = f (u0, v)

+∞

−∞ dv f (v, u0)

slide-6
SLIDE 6

Let u and v be two Cartesian parameters (then, volumetric probabilities and probability densities are identical). Given a probability density f (u, v) , one defines the two marginal probability densities fu(u) =

+∞

−∞ dv f (u, v)

fv(v) =

+∞

−∞ du f (u, v)

and the two conditional probability densities fu|v(u|v) = f (u, v)

+∞

−∞ du f (u, v)

fv|u(v|u) = f (u, v)

+∞

−∞ dv f (v, u)

slide-7
SLIDE 7

One has fu|v(u|v) = f (u, v) fv(v) fv|u(v|u) = f (u, v) fu(u) from where (a joint distribution can be expressed by a condi- tional distribution times a marginal distribution) f (u, v) = fu|v(u|v) fv(v) = fv|u(v|u) fu(u) from where (Bayes theorem) fu|v(u|v) = fv|u(v|u) fu(u) fv(v)

slide-8
SLIDE 8

Recall: f (u, v) = fu|v(u|v) fv(v) = fv|u(v|u) fu(u) . The two quantities u and v are said to have independent un- certainties if, in fact, f (u, v) = fu(u) fv(v) (the joint distribution equals the product of the two marginal distributions). This implies (and is implied by) fu|v(u|v) = fu(u) ; fv|u(v|u) = fv(v) .

slide-9
SLIDE 9
  • 5

5 10 15 20

  • 10
  • 5

5 10

  • 5

5 10 15 20

  • 10
  • 5

5 10

two quantities with independent uncertainties (the joint distribution is the product

  • f the two marginal distributions)
slide-10
SLIDE 10

Let u and v be two Cartesian parameters (then, volumet- ric probabilities and probability densities are identical). Let f (u, v) , be a probability density that is not qualitatively differ- ent from a two-dimensional Gaussian. The mean values are u =

+∞

−∞ du

+∞

−∞ dv u f (u, v)

v =

+∞

−∞ du

+∞

−∞ dv v f (u, v)

the variances are cuu = σ2

u =

+∞

−∞ du

+∞

−∞ dv (u − u)2 f (u, v)

cvv = σ2

v =

+∞

−∞ du

+∞

−∞ dv (v − v)2 f (u, v)

and the covariance is cuv =

+∞

−∞ du

+∞

−∞ dv (u − u)(v − v) f (u, v)

slide-11
SLIDE 11

The covariance matrix is C =

  • cuu

cuv cvu cvv

  • =
  • σ2

u

cuv cvu σ2

v

  • .

It is symmetric and positive definite (or, at least, non-negative). Note: the correlation, defined as ρuv = cuv σu σv

=

cuv

√cuu √cvv

, has the property

−1 ≤ ρuv ≤ +1 .

slide-12
SLIDE 12

The general form of a covariance matrix is C =        c11 c12 c13 . . . c21 c22 c23 . . . c31 c32 c33 . . . . . . . . . . . . ...       

=

       σ2

1

c12 c13 . . . c21 σ2

2

c23 . . . c31 c32 σ2

3

. . . . . . . . . . . . ...        . The quantities with immediate interpretation are the standard deviations

{σ1, σ2, σ3, . . . }

and the correlation matrix R =        1 ρ12 ρ13 . . . ρ21 1 ρ23 . . . ρ31 ρ32 1 . . . . . . . . . . . . ...       

slide-13
SLIDE 13

The multidimensional Gaussian distribution is defined as f (x1, x2, . . . , xn) ≡ f (x) = k exp − 1

2 (x − x0)t C-1 (x − x0) )

Its mean is x0 and its covariance is C (not obvious!).