Functional Equations & Neural Networks for Time Series Interpolation
Lars Kindermann, AWI Achim Lewandowski, OEFAI
Drop an object with different speeds and measure the speed at the ground
Question: What’s the speed
v
m after half the way at xm?v0 v1
x0
x1
v1 = f v 0 vm = ? xm
v0
v0 v1
free fall
friction Data
An old experiment
Free Fall:
Theory:
Model: with data fitted
With additional Friction:
Theory:
Model: Integrate numerically and fit and - already a non-trivial Problem!
t2
2
x = g
v1 = f v 0 = v02 + 2g x g
t2
2
x g k1
t
– x k2
t
x 2
– f
t
x –
=
g k
Solving with traditional Physics
x0
x1
v1 v0
x0
x1
vm = vm xm
v0
v1 = f v 0
v1 = vm = v0
divide into two equal steps...
v = f v
and solve this functional equation for Theory: Assume translation invariance
A Data-based Aproach
A solution of this equation is a kind of square root of the function .
•
If : is a function, we look for another function which composed with itself equals :Because the self-composition of a function is also called
“iteration”, the square root of a function is usually called its iterative root.
is solved by the fractional iterates of a function :
x = f x
f
f x IRn IRn x
f x = f x
f f x = f2 x
n x = fm x f
x = fm n x
A Functional Equation
A solution of this equation is called a square root of .
•
If : is a function, we look for another function which composed with itself equals :Because the self-composition of a function is also called “iteration”, the square root of a function is usually called its iterative root.
is solved by the fractional iterates of a function :
x = f x
f
f x IRn IRn x
f x = f x
f f x = f2 x
n x = fm x f
x = fm n x
= f
x
x
y y
= f
x
x
y
y
f
3 5---
A Functional Equation
The exponential notation of the iteration of functions can be extended beyond integer exponents:
•
means•
for positive integers are the well known iterations of•
denotes the identity function,•
is the inverse funktion of•
is the -th iteration of the inverse of•
is the -th iterative root of•
is the -th iteration of the -th iterative root or fractional iterate of The family forms the continuous iteration group of .Within this the translation equation is satisfied.
fn x
f 1 f
f n n f
f0 f0 x = x
f –1 f
f –n n f
f 1 n n f
fm n m n f
ft x f
f a b+ x = f af b x
Generalized Iteration
Map this to a Network
x
x
f x
f
share weights
f1 n x
= fm n
f x
1 m n
x f x
loop n times
•
Weight Copy: Train only the last layer and copy the weights continously backwards•
Weight Sharing: Initialize corresponding weights with equal values and sum up all delivered by the network learning rule•
Weight Coupling: Start with different values and let the corresponding weights of the iteration layers approach each other by a term like•
Regularization: Add a penalty term to the error function which assigns an error to the weight-differences to regularize the network. This allows to uti- lize second order gradient methods like quasi Newton for faster training.•
Exact Gradient: Compute the exact gradients for an iterated Network wi
wi
= wj – wi
Training Methods
0 1 2 3 4 5 6 7 8 9 10 0
1 2 3 4 5 6 7 8 9 10
Start Velocity [m/s]
End Velocity [m/s]
measurement for v1 (training data) physics for vm (prediction task) fractional iterates (network results)
v1 = f v 0
f 0 v0 = v0 vm = f 1 2 v0 f 1 4
f3 4
0 1 2 3 4 5 6 7 8 9 10
0 1 2 3 4 5 6 7 8 9 10
Start Velocity [m/s]
End Velocity [m/s]
measurement for v1 (training data) physics for vm (prediction task) fractional iterates (network results)
v1 = f v 0
vm = f 1 2 v0 f 0 v0 = v0
f3 4 f 1 4
no friction with friction
The Network results are conform to the laws of physics up to a mean error of 10-6
Results for „The Fall“
0
0.2
0.4
0.6
0.8
1
1.2
0 1 2 3 4 5 6 7 8 9 10
0 2 4 6 8 10 12
h [m]
v0 [m/s]
v [m/s]
Training Data
trajectories iterative roots
v
0v(h)
v=f(v
0,h)
h
The Embedding Problem
One of the most important functional equations:
The Eigenvalue problem of functional calculus.
Transform to:
f x = c x
f x = –1c x invert
–1
x
x
f x c
train
f
The Schröder Equation
Commuting Functions
f x
f x x
x
f
f weight
sharing
outputs targets
x
f x = f x
-
The steel bands are processed by N identical stands in a row-
, are known and can be measured-
x2
x1 xout
f x in,p1 f x 1,p2
F x in,p1pN
f x N 1– ,pN
xin
Measuring
p1 p2 pN
instrument
=set of parameters like force, heat, strip thikness and width...
pi
xin pi xout
xout = F x in p1...pN = f ...f f x in p1 p2... p, N
Steel Mill Model
Steel Mill Network
For a given autoregressive Box-Jenkins AR(n) timeseries , we
define the function : which maps the vector of the last n samples
one step into the future as
and can simply write now.
The discrete time evolution of the the system can be calculated using the
matrix powers of F: .
xt akxt k–
k = 1 n
=
F Rn Rn
xt 1– = xt 1– xt n– xt = xt xt 1– xt– n 1–
F
a1 a2 an 1 0 0 0 0 1 0 0 0 0 1 0
= xt = F x t 1–
xt n+ = Fn xt 1–
Timeseries Interpolation
This autoregressive system is called linear embeddable if the matrix power exists also for all real . This is the case if can be decomposed into
with being a diagonal matrix consisting of the eigenvalues of and being an invertible square matrix which columns are the eigenvec- tors of . Additionally all must be non-negative to have a linear and real embedding, otherwise we will get a complex embedding.
Then we can obtain with
Now we have a continuous function and the interpolation of the original time series consists of the first element of .
Ft
t R+ F
F = S A S –1 A i
F S
F i
Ft = S A t S–1 At
1t 0 0
0 0
0 0 nt
=
x t = Ft x0
x t x
Using Generalized Matrix Powers
The Fibonacci series , , is generated by and . By eigenvalue decomposition of we get
That is Binet’s formula in the first component
x0 = 0 x1 = 1 xt = xt 1– + xt 2–
F 1 1
= 1 0 x1 = 1 0 F
xt 1+ = Ftx1 = SAtS–1x1
1+ 5
--- 12 – 5 ---2
1 1
1+ 5
---2
t
0 0 1---–2 5t
1
---5 1
2--- 1
2 5
--- –
1
5
--- –
1
2--- 1
2 5
---
+
1
= 0
xt 1
--- 15 + 5 ---2
t 1 – 5
---2
t
–
=
Example: A continuous Fibonacci Function
A time series of yearly snapshots from a discrete non linear Lotka-Volterra type predator - prey system (x = hare, y = lynx) is used as training data:
and
From these samples we calculate the monthly population by use of a neural network based method to compute iterative roots and fractional iterates.
The given method provides a natural way to estimate not only the values over a year, but also to extrapolate arbitrarily smooth into the future.
xt 1+ = 1 a b y+ – t xt yt 1+ = 1 c– + d xt yt
Nonlinear Example
0 2 4 6 8 10 12
0 1 2 3 4 5 6 7
Prey
Predator