MATH 414 Lecture 2

From Notes
Jump to navigation Jump to search

« previous | Wednesday, January 15, 2014 | next »


Inner Product

Most common example is the vector dot product:

⟨u→,v→⟩=u→⋅v→=∑iuivi=|u→||v→|cos⁡θ

Generalized Algebraic Structure

We induce a geometry on the algebraic definition

Length: ‖v→‖=⟨v→,v→⟩

Angle Between: θ=cos−1(⟨u→,v→⟩‖u→‖‖v→‖)

The most important case is where θ=π2, in which case cos⁡θ=0 and—more importantly—the vectors u→ and v→ are perperdicular.

Properties of the Inner Product

Positivity: ⟨u→,u→⟩>0 unless u→=0→, in which case ⟨0→,0→=0⟩

Conjugate Symmetry: ⟨u→,v→⟩=⟨v→,u→⟩‾. In the case of real numbers, the property becomes commutativity: ⟨u→,v→⟩=⟨v→,u→⟩

Homogenuity: ⟨cu→,v→⟩=c⟨u→,v→⟩

Linearity: ⟨u→+w→,v→⟩=⟨u→,v→⟩+⟨w→,v→⟩

Real Spaces

for ℝn, where x→=[x1⋮xn] and y→=[y1⋮yn], we have

⟨x→,y→⟩=yTx=∑k=1nxkyk


Complex Spaces

for ℂn, ⟨x→,y→⟩ is given by y→*x→ (also written y→Hx→), where y→*=y→H is the conjugate transpose of y→.

Example: y→=[i1−1+2i3+i], y→*=[−i1−1−2i3−i]

Some concrete implementations of inner products may not satisfy all properties for all members of the space. [1]


Signal Spaces

Continuous and Discrete

A signal is just f(t) measured on [a,b].

Signals and Circuits

For current I(t) (essentially our function f(t))

We can measure voltage drop of a signal V(t)=RI(t)

Power at time t is given by P(t)=VI(t)=RI2(t)

Energy in time interval [a,b] is given by E=∫abP(t)dt=R∫ab|I(t)|2dt.

Hence E=∫ab|f(t)|2dt

Finite Energy Signals

Complex valued f(t) gives amplitude and phase.


Finite energy: ∫ab|f(t)|2dt<∞. Moreover, ∫∞∞|f(t)|2dt<∞

Signals don't need to be continuous (digital circuit signals) [2]

Continuous Signal Space

All signals with finite energy fit into a space:

L2={f:[a,b]→ℂ∣∫ab|f(t)|2dt<∞}

The L comes from mathematician Lebesgue (ca. 1900), and the "square" comes from the square in the integral. The reason for this name is because integral we compute is a Lebesgue integral, not a Riemann integral. [3]

We define an inner product on L2 as

⟨f,g⟩L2[a,b]=∫abf(t)g(t)‾dt


Discrete Signal Space

Notation: l2

By sampling the function at fixed points, we obtain a sequence X={xn}n=−∞∞=(…,x−1,x0,x1,…)

Thus l2⊂L2.

We define our inner product by a discrete sum:

⟨X,Y⟩l2=∑n=−∞∞xny¯n

The energy in a discrete signal is similarly given by

∑n=−∞∞|xn|2


Inner Product Spaces

Schwarz's Inequality

Using axioms from definition, we arrive at

|⟨u→,v→⟩|≤‖u→‖‖u→‖, where ‖u→‖=⟨u→,u→⟩ represents the norm of a vector u→∈V ("LENGTH")

Triangle Inequality

Property of length:

‖u→+v→‖≤‖u→‖+‖v→‖.

"The sum of the lengths any two sides of a triangle must be at least the length of the third side."

Footnotes

  1. ↑ 1700s and Black Swan
  2. ↑ static pops are just discontinuities; static noise is a bunch of discontinuities in the signal
  3. ↑ A Lebesgue integral integrates over the y axis instead of the x axis. This may sound like a dumb idea, but it's brilliant because it allows us to integrate functions that would otherwise not be integrable.