Slow Feature Analysis is an unsupervised learning algorithm that aims to learn invariant or at least slowly changing features from a vector signal. It is based on the major axis transformation .
Problem Description
If an input signal is given, an input / output function is sought for which varies as little as possible and is not constant.
x
(
t
)
{\ displaystyle x (t)}
G
(
x
)
{\ displaystyle g (x)}
y
(
t
)
=
G
(
x
(
t
)
)
{\ displaystyle y (t) = g (x (t))}
G
{\ displaystyle g}
Formally one writes:
A -dimensional input signal is given with . Find a -dimensional input / output function that generates the -dimensional output with for each . The following constraints must be met for all of them :
n
{\ displaystyle n}
x
(
t
)
=
[
x
1
(
t
)
,
...
,
x
n
(
t
)
]
{\ displaystyle x (t) = [x_ {1} (t), \ dots, x_ {n} (t)]}
t
∈
[
t
0
,
t
1
]
{\ displaystyle t \ in [t_ {0}, t_ {1}]}
m
{\ displaystyle m}
G
(
x
)
=
[
G
1
(
x
)
,
...
,
G
m
(
x
)
]
{\ displaystyle g (x) = [g_ {1} (x), \ dots, g_ {m} (x)]}
x
(
t
)
{\ displaystyle x (t)}
m
{\ displaystyle m}
y
(
t
)
=
[
y
1
(
t
)
,
y
2
(
t
)
,
...
,
y
m
(
t
)
]
{\ displaystyle y (t) = [y_ {1} (t), y_ {2} (t), \ dots, y_ {m} (t)]}
y
i
(
t
)
=
G
i
(
x
(
t
)
)
{\ displaystyle y_ {i} (t) = g_ {i} (x (t))}
i
∈
{
1
,
...
,
m
}
{\ displaystyle i \ in \ {1, \ dots, m \}}
i
∈
{
1
,
...
,
m
}
{\ displaystyle i \ in \ {1, \ dots, m \}}
Δ
i
=
Δ
(
y
i
)
:
=
⟨
y
˙
i
2
⟩
is minimal
⟨
y
i
⟩
=
0
(Average)
⟨
y
i
2
⟩
=
1
(Variance)
∀
i
′
<
i
:
⟨
y
i
′
y
i
⟩
=
0
(Decorrelation)
{\ displaystyle {\ begin {aligned} \ Delta _ {i} = \ Delta (y_ {i}): & = \ langle {\ dot {y}} _ {i} ^ {2} \ rangle & {\ text {is minimal}} \\\ langle y_ {i} \ rangle & = 0 & {\ text {(mean)}} \\\ langle y_ {i} ^ {2} \ rangle & = 1 & {\ text {(variance )}} \\\ forall i '<i: \ langle y_ {i'} y_ {i} \ rangle & = 0 & {\ text {(decorrelation)}} \ end {aligned}}}
where the derivative is denoted by and an average over time is:
y
˙
{\ displaystyle {\ dot {y}}}
t
{\ displaystyle t}
⟨
⋅
⟩
{\ displaystyle \ langle \ cdot \ rangle}
⟨
f
⟩
: =
1
t
1
-
t
0
∫
t
0
t
1
f
(
t
)
d
t
{\ displaystyle \ langle f \ rangle: = {\ frac {1} {t_ {1} -t_ {0}}} \ int \ limits _ {t_ {0}} ^ {t_ {1}} f (t) \ mathrm {d} t}
Web links
Individual evidence
↑ Laurenz Wiskott, Terrence J. Sejnowski: Slow Feature Analysis: Unsupervised Learning of Invariances . In: Neural Computation . tape 14 , no. 4 , 2002, p. 715-770 , doi : 10.1162 / 089976602317318938 .
<img src="https://de.wikipedia.org/wiki/Special:CentralAutoLogin/start?type=1x1" alt="" title="" width="1" height="1" style="border: none; position: absolute;">