Machine Learning:

Perceptrons
Prof. Dr. Martin Riedmiller

Albert-Ludwigs-University Freiburg
AG Maschinelles Lernen

Machine Learning: Perceptrons p.1/24

 Neural Networks
The human brain has approximately 1011 neurons
Switching time 0.001s (computer 1010 s)
Connections per neuron: 104 105
0.1s for face recognition
I.e. at most 100 computation steps
parallelism
additionally: robustness, distributedness
ML aspects: use biology as an inspiration for artificial neural models and
algorithms; do not try to explain biology: technically imitate and exploit
capabilities

Machine Learning: Perceptrons p.2/24

 Biological Neurons
Dentrites input information to the cell
Neuron fires (has action potential) if a certain threshold for the voltage is
exceeded
Output of information by axon
The axon is connected to dentrites of other cells via synapses
Learning corresponds to adaptation of the efficiency of synapse, of the
synaptical weight
dendrites
SYNAPSES

AXON

soma

Machine Learning: Perceptrons p.3/24

 19
42
a
19 rtifi
49 ci

1950
H al n
eb eu
bi
an ron
le s (M
19 ar
5
19 8
ni
ng
cC
R ul

1960
1960 os (H loc
60 Ad en eb h/
a
Le lin la b b )
Pi
tts
rn e/ tt p )
19 m M e
6 at Ad rc
19 9 rix a ep
7 p (S
lin tr

1970
t e on
19 0 er ei (W (
72 ev ce
o p nb i R
se lut tr d
uc ro os
lf- ion on h) w en
or a s /H bl
ga ry (M of att
ni al in f) )
zi go s

1980
19 ng r ky
82 m ithm/Pa
H ap s p
19 op s (R ert
86 fie (K e )
oh ch
Ba l d n on en
ck et en be
p w
1990

19 ro ) rg
pa
or
ks )
co 92
m Ba ga (H
su p y t io op
p ut es n fie
Bo po at i
i n (o ld
os rt on fe r i g )
tin ve al re .
2000

g cto lea nc 19
r m rn e 74
ac ing )
Historical ups and downs

hi th
Machine Learning: Perceptrons p.4/24

ne e
s ory
 Perceptrons: adaptive neurons
perceptrons (Rosenblatt 1958, Minsky/Papert 1969) are generalized variants
of a former, more simple model (McCulloch/Pitts neurons, 1942):
inputs are weighted
weights are real numbers (positive and negative)
no special inhibitory inputs

Machine Learning: Perceptrons p.5/24

 Perceptrons: adaptive neurons
perceptrons (Rosenblatt 1958, Minsky/Papert 1969) are generalized variants
of a former, more simple model (McCulloch/Pitts neurons, 1942):
inputs are weighted
weights are real numbers (positive and negative)
no special inhibitory inputs
a percpetron with n inputs is described by a weight vector
~ = (w1 , . . . , wn )T Rn and a threshold R. It calculates the
w
following function:
(
T 1 if x1 w1 + x2 w2 + + xn wn
(x1 , . . . , xn ) 7 y=
0 if x1 w1 + x2 w2 + + xn wn <

Machine Learning: Perceptrons p.5/24

 Perceptrons: adaptive neurons
(cont.)
for convenience: replacing the threshold by an additional weight (bias weight)
w0 = . A perceptron with weight vector w ~ and bias weight w0 performs
the following calculation:

n
X
(x1 , . . . , xn )T 7 y = fstep (w0 + (wi xi )) = fstep (w0 + hw,
~ ~xi)
i=1

with
(
1 if z 0
fstep (z) =
0 if z < 0

Machine Learning: Perceptrons p.6/24

 Perceptrons: adaptive neurons
(cont.)
for convenience: replacing the threshold by an additional weight (bias weight)
w0 = . A perceptron with weight vector w ~ and bias weight w0 performs
the following calculation:

n
X
(x1 , . . . , xn )T 7 y = fstep (w0 + (wi xi )) = fstep (w0 + hw,
~ ~xi)
i=1

with
(
1 if z 0
fstep (z) = 1
0 if z < 0
w0
x1 w1
.. y
.
xn wn

Machine Learning: Perceptrons p.6/24

 Perceptrons: adaptive neurons
(cont.)
geometric interpretation of a x2
upper
perceptron: halfspace

input patterns (x1 , . . . , xn ) are

points in n-dimensional space hyp
erp
lane
lower
halfspace

x1
x3

upper
halfspace

lower
halfspace x2

hy
pe
rpl
an
e
x1

Machine Learning: Perceptrons p.7/24

 Perceptrons: adaptive neurons
(cont.)
geometric interpretation of a x2
upper
perceptron: halfspace

input patterns (x1 , . . . , xn ) are

points in n-dimensional space hyp
erp
lane
points with w0 + hw, ~ ~xi = 0 are on lower
halfspace
a hyperplane defined by w0 and w ~ x1
x3

upper
halfspace

lower
halfspace x2

hy
pe
rpl
an
e
x1

Machine Learning: Perceptrons p.7/24

 Perceptrons: adaptive neurons
(cont.)
geometric interpretation of a x2
upper
perceptron: halfspace

input patterns (x1 , . . . , xn ) are

points in n-dimensional space hyp
erp
lane
points with w0 + hw, ~ ~xi = 0 are on lower
halfspace
a hyperplane defined by w0 and w ~ x1
points with w0 + hw, ~ ~xi > 0 are x3

upper
above the hyperplane halfspace

lower
halfspace x2

hy
pe
rpl
an
e
x1

Machine Learning: Perceptrons p.7/24

 Perceptrons: adaptive neurons
(cont.)
geometric interpretation of a x2
upper
perceptron: halfspace

input patterns (x1 , . . . , xn ) are

points in n-dimensional space hyp
erp
lane
points with w0 + hw, ~ ~xi = 0 are on lower
halfspace
a hyperplane defined by w0 and w ~ x1
points with w0 + hw, ~ ~xi > 0 are x3

upper
above the hyperplane halfspace

points with w0 + hw,

~ ~xi < 0 are
below the hyperplane
lower
halfspace x2

hy
pe
rpl
an
e
x1

Machine Learning: Perceptrons p.7/24

 Perceptrons: adaptive neurons
(cont.)
geometric interpretation of a x2
upper
perceptron: halfspace

input patterns (x1 , . . . , xn ) are

points in n-dimensional space hyp
erp
lane
points with w0 + hw, ~ ~xi = 0 are on lower
halfspace
a hyperplane defined by w0 and w ~ x1
points with w0 + hw, ~ ~xi > 0 are x3

upper
above the hyperplane halfspace

points with w0 + hw,

~ ~xi < 0 are
below the hyperplane
lower

perceptrons partition the input space halfspace x2

hy
pe
into two halfspaces along a rpl
an
e
hyperplane x1

Machine Learning: Perceptrons p.7/24

 Perceptron learning problem
perceptrons can automatically adapt to example data Supervised
Learning: Classification

Machine Learning: Perceptrons p.8/24

 Perceptron learning problem
perceptrons can automatically adapt to example data Supervised
Learning: Classification
perceptron learning problem:
given:
a set of input patterns P Rn , called the set of positive examples
another set of input patterns N Rn , called the set of negative
examples
task:
generate a perceptron that yields 1 for all patterns from P and 0 for all
patterns from N

Machine Learning: Perceptrons p.8/24

 Perceptron learning problem
perceptrons can automatically adapt to example data Supervised
Learning: Classification
perceptron learning problem:
given:
a set of input patterns P Rn , called the set of positive examples
another set of input patterns N Rn , called the set of negative
examples
task:
generate a perceptron that yields 1 for all patterns from P and 0 for all
patterns from N
obviously, there are cases in which the learning task is unsolvable, e.g.
P N = 6

Machine Learning: Perceptrons p.8/24

 Perceptron learning problem
(cont.)
Lemma (strict separability):
Whenever exist a perceptron that classifies all training patterns accurately,
there is also a perceptron that classifies all training patterns accurately and
no training pattern is located on the decision boundary, i.e.
w~0 + hw,~ ~xi =
6 0 for all training patterns.

Machine Learning: Perceptrons p.9/24

 Perceptron learning problem
(cont.)
Lemma (strict separability):
Whenever exist a perceptron that classifies all training patterns accurately,
there is also a perceptron that classifies all training patterns accurately and
no training pattern is located on the decision boundary, i.e.
w~0 + hw,~ ~xi =
6 0 for all training patterns.
Proof:
Let (w,
~ w0 ) be a perceptron that classifies all patterns accurately. Hence,
(
0 for all ~
x P
hw,
~ ~xi + w0
<0 xN
for all ~

Machine Learning: Perceptrons p.9/24

 Perceptron learning problem
(cont.)
Lemma (strict separability):
Whenever exist a perceptron that classifies all training patterns accurately,
there is also a perceptron that classifies all training patterns accurately and
no training pattern is located on the decision boundary, i.e.
w~0 + hw,~ ~xi =
6 0 for all training patterns.
Proof:
Let (w,
~ w0 ) be a perceptron that classifies all patterns accurately. Hence,
(
0 for all ~
x P
hw,
~ ~xi + w0
<0 xN
for all ~
Define = min{(hw,~ ~xi + w0 )|~x N }. Then:
(
2 > 0 xP
for all ~
hw,
~ ~xi + w0 +
2 2 < 0 for all ~x N

Machine Learning: Perceptrons p.9/24

 Perceptron learning problem
(cont.)
Lemma (strict separability):
Whenever exist a perceptron that classifies all training patterns accurately,
there is also a perceptron that classifies all training patterns accurately and
no training pattern is located on the decision boundary, i.e.
w~0 + hw,~ ~xi =
6 0 for all training patterns.
Proof:
Let (w,
~ w0 ) be a perceptron that classifies all patterns accurately. Hence,
(
0 for all ~
x P
hw,
~ ~xi + w0
<0 xN
for all ~
Define = min{(hw,~ ~xi + w0 )|~x N }. Then:
(
2 > 0 xP
for all ~
hw,
~ ~xi + w0 +
2 2 < 0 for all ~x N
Thus, the perceptron (w,
~ w0 + 2 ) proves the lemma.
Machine Learning: Perceptrons p.9/24
 Perceptron learning algorithm:
idea
assume, the perceptron makes an
x P:
error on a pattern ~
hw,
~ ~xi + w0 < 0
how can we change w
~ and w0 to
avoid this error?

Machine Learning: Perceptrons p.10/24

 Perceptron learning algorithm:
idea
assume, the perceptron makes an
x P:
error on a pattern ~
hw,
~ ~xi + w0 < 0
how can we change w
~ and w0 to
avoid this error? we need to
increase hw,
~ ~xi + w0

Machine Learning: Perceptrons p.10/24

 Perceptron learning algorithm:
idea
assume, the perceptron makes an
x P:
error on a pattern ~
hw,
~ ~xi + w0 < 0
how can we change w
~ and w0 to
avoid this error? we need to
increase hw,
~ ~xi + w0
increase w0
if xi > 0, increase wi
if xi < 0 (negative influence),
decrease wi
perceptron learning algorithm: add ~x
~ , add 1 to w0 in this case. Errors
to w
on negative patterns: analogously.

Machine Learning: Perceptrons p.10/24

 Perceptron learning algorithm:
idea
assume, the perceptron makes an
x P:
error on a pattern ~
hw,
~ ~xi + w0 < 0
how can we change w
~ and w0 to
avoid this error? we need to
increase hw,
~ ~xi + w0
increase w0
if xi > 0, increase wi
if xi < 0 (negative influence),
decrease wi
perceptron learning algorithm: add ~x
~ , add 1 to w0 in this case. Errors
to w
on negative patterns: analogously.

Machine Learning: Perceptrons p.10/24

 Perceptron learning algorithm:
idea
assume, the perceptron makes an x3
x P:
error on a pattern ~
hw,
~ ~xi + w0 < 0
how can we change w
~ and w0 to
avoid this error? we need to w

increase hw,
~ ~xi + w0 x2

increase w0
if xi > 0, increase wi x1
if xi < 0 (negative influence),
decrease wi
Geometric intepretation: increasing w0
perceptron learning algorithm: add ~x
~ , add 1 to w0 in this case. Errors
to w
on negative patterns: analogously.

Machine Learning: Perceptrons p.10/24

 Perceptron learning algorithm:
idea
assume, the perceptron makes an x3
x P:
error on a pattern ~
hw,
~ ~xi + w0 < 0
how can we change w
~ and w0 to
avoid this error? we need to w

increase hw,
~ ~xi + w0 x2

increase w0
if xi > 0, increase wi x1
if xi < 0 (negative influence),
decrease wi
Geometric intepretation: increasing w0
perceptron learning algorithm: add ~x
~ , add 1 to w0 in this case. Errors
to w
on negative patterns: analogously.

Machine Learning: Perceptrons p.10/24

 Perceptron learning algorithm:
idea
assume, the perceptron makes an x3
x P:
error on a pattern ~
hw,
~ ~xi + w0 < 0
how can we change w
~ and w0 to
avoid this error? we need to
increase hw,
~ ~xi + w0 w x2

increase w0
if xi > 0, increase wi x1
if xi < 0 (negative influence),
decrease wi
Geometric intepretation: increasing w0
perceptron learning algorithm: add ~x
~ , add 1 to w0 in this case. Errors
to w
on negative patterns: analogously.

Machine Learning: Perceptrons p.10/24

 Perceptron learning algorithm:
idea
assume, the perceptron makes an
x P:
error on a pattern ~
hw,
~ ~xi + w0 < 0
how can we change w
~ and w0 to
avoid this error? we need to
increase hw,
~ ~xi + w0
increase w0
if xi > 0, increase wi
if xi < 0 (negative influence),
decrease wi
perceptron learning algorithm: add ~x
~ , add 1 to w0 in this case. Errors
to w
on negative patterns: analogously.

Machine Learning: Perceptrons p.10/24

 Perceptron learning algorithm:
idea
assume, the perceptron makes an x3
x P:
error on a pattern ~
hw,
~ ~xi + w0 < 0
how can we change w
~ and w0 to
avoid this error? we need to w

increase hw,
~ ~xi + w0 x2

increase w0
if xi > 0, increase wi x1
if xi < 0 (negative influence),
decrease wi
Geometric intepretation: modifying w
~
perceptron learning algorithm: add ~x
~ , add 1 to w0 in this case. Errors
to w
on negative patterns: analogously.

Machine Learning: Perceptrons p.10/24

 Perceptron learning algorithm:
idea
assume, the perceptron makes an x3
x P:
error on a pattern ~
hw,
~ ~xi + w0 < 0
how can we change w
~ and w0 to w

avoid this error? we need to w

increase hw,
~ ~xi + w0 x2

increase w0
if xi > 0, increase wi x1
if xi < 0 (negative influence),
decrease wi
Geometric intepretation: modifying w
~
perceptron learning algorithm: add ~x
~ , add 1 to w0 in this case. Errors
to w
on negative patterns: analogously.

Machine Learning: Perceptrons p.10/24

 Perceptron learning algorithm:
idea
assume, the perceptron makes an x3
x P:
error on a pattern ~
hw,
~ ~xi + w0 < 0
how can we change w
~ and w0 to w

avoid this error? we need to

increase hw,
~ ~xi + w0 x2

increase w0
if xi > 0, increase wi x1
if xi < 0 (negative influence),
decrease wi
Geometric intepretation: modifying w
~
perceptron learning algorithm: add ~x
~ , add 1 to w0 in this case. Errors
to w
on negative patterns: analogously.

Machine Learning: Perceptrons p.10/24

 Perceptron learning algorithm
Require: positive training patterns P and a negative training examples N
Ensure: if exists, a perceptron is learned that classifies all patterns accurately

1: initialize weight vector w

~ and bias weight w0 arbitrarily
2: x P N do
while exist misclassified pattern ~
3: x P then
if ~
4: ~ w
w ~ + ~x
5: w0 w0 + 1
6: else
7: ~ w
w ~ ~x
8: w0 w0 1
9: end if
10: end while
11: return w ~ and w0

Machine Learning: Perceptrons p.11/24

 Perceptron learning algorithm:
example

N = {(1, 0)T , (1, 1)T }, P = {(0, 1)T }

exercise

Machine Learning: Perceptrons p.12/24

 Perceptron learning algorithm:
convergence
Lemma (correctness of perceptron learning):
Whenever the perceptron learning algorithm terminates, the perceptron
given by (w,
~ w0 ) classifies all patterns accurately.

Machine Learning: Perceptrons p.13/24

 Perceptron learning algorithm:
convergence
Lemma (correctness of perceptron learning):
Whenever the perceptron learning algorithm terminates, the perceptron
given by (w,
~ w0 ) classifies all patterns accurately.
Proof: follows immediately from algorithm.

Machine Learning: Perceptrons p.13/24

 Perceptron learning algorithm:
convergence
Lemma (correctness of perceptron learning):
Whenever the perceptron learning algorithm terminates, the perceptron
given by (w,
~ w0 ) classifies all patterns accurately.
Proof: follows immediately from algorithm.
Theorem (termination of perceptron learning):
Whenever exists a perceptron that classifies all training patterns correctly,
the perceptron learning algorithm terminates.

Machine Learning: Perceptrons p.13/24

 Perceptron learning algorithm:
convergence
Lemma (correctness of perceptron learning):
Whenever the perceptron learning algorithm terminates, the perceptron
given by (w,
~ w0 ) classifies all patterns accurately.
Proof: follows immediately from algorithm.
Theorem (termination of perceptron learning):
Whenever exists a perceptron that classifies all training patterns correctly,
the perceptron learning algorithm terminates.
Proof:
for simplification we will add the bias weight to the weight vector, i.e.
w~ = (w0 , w1 , . . . , wn )T , and 1 to all patterns, i.e. ~x = (1, x1 , . . . , xn )T .
~ (t) the weight vector in the t-th iteration of perceptron
We will denote with w
x(t) the pattern used in the t-th iteration.
learning and with ~

Machine Learning: Perceptrons p.13/24

 Perceptron learning algorithm:
convergence proof (cont.)
Let be w
~ a weight vector that strictly classifies all training patterns.

Machine Learning: Perceptrons p.14/24

 Perceptron learning algorithm:
convergence proof (cont.)
Let be w
~ a weight vector that strictly classifies all training patterns.
(t+1) (t) (t)

w
~ ,w
~ = w ~ ~x
~ ,w

(t) (t)
= w
~ ,w
~ w~ , ~x
(t)
w
~ ,w
~ +

with ~ , ~xi |~x P} { hw

:= min ({hw ~ , ~xi |~x N })

Machine Learning: Perceptrons p.14/24

 Perceptron learning algorithm:
convergence proof (cont.)
Let be w
~ a weight vector that strictly classifies all training patterns.
(t+1) (t) (t)

w
~ ,w
~ = w ~ ~x
~ ,w

(t) (t)
= w
~ ,w
~ w~ , ~x
(t)
w
~ ,w
~ +

with := min ({hw ~ , ~xi |~x P} { hw ~ , ~xi |~x N })

> 0 since w
~ strictly classifies all patterns

Machine Learning: Perceptrons p.14/24

 Perceptron learning algorithm:
convergence proof (cont.)
Let be w
~ a weight vector that strictly classifies all training patterns.
(t+1) (t) (t)

w
~ ,w
~ = w ~ ~x
~ ,w

(t) (t)
= w
~ ,w
~ w~ , ~x
(t)
w
~ ,w
~ +

with := min ({hw ~ , ~xi |~x P} { hw ~ , ~xi |~x N })

> 0 since w
~ strictly classifies all patterns
Hence,
(t+1) (0)

w
~ ,w
~ w
~ ,w
~
+ (t + 1)

Machine Learning: Perceptrons p.14/24

 Perceptron learning algorithm:
convergence proof (cont.)

(t+1) 2 (t+1) (t+1)

||w
~ || = w ~ ,w
~
(t) (t) (t) (t)

= w ~ ~x , w ~ ~x
(t) 2
(t) (t)
= ||w
~ || 2 w ~ , ~x + ||~x(t) ||2
~ (t) ||2 +
||w

with := max{||~x||2 |~x P N }

Machine Learning: Perceptrons p.15/24

 Perceptron learning algorithm:
convergence proof (cont.)

(t+1) 2 (t+1) (t+1)

||w
~ || = w ~ ,w
~
(t) (t) (t) (t)

= w ~ ~x , w ~ ~x
(t) 2
(t) (t)
= ||w
~ || 2 w ~ , ~x + ||~x(t) ||2
~ (t) ||2 +
||w

with := max{||~x||2 |~x P N }

Hence,
~ (t+1) ||2 ||w
||w ~ (0) ||2 + (t + 1)

Machine Learning: Perceptrons p.15/24

 Perceptron learning algorithm:
convergence proof (cont.)

(t+1)

(t+1) w~ ,w
~
cos (w

~ ,w
~ )=
||w ~ (t+1) ||
~ || ||w

Machine Learning: Perceptrons p.16/24

 Perceptron learning algorithm:
convergence proof (cont.)

(t+1)

(t+1) w~ ,w ~
cos (w

~ ,w
~ )=
||w ~ (t+1) ||
~ || ||w

(0)

w
~ ,w
~ + (t + 1)

p
~ (0) ||2 + (t + 1)
~ || ||w
||w

Machine Learning: Perceptrons p.16/24

 Perceptron learning algorithm:
convergence proof (cont.)

(t+1)

(t+1) w~ ,w ~
cos (w
~ ,w
~ )=
||w ~ (t+1) ||
~ || ||w

(0)

w
~ ,w
~
+ (t + 1)
p
~ || ||w
||w ~ (0) ||2 + (t + 1) t

Since cos (w ~ (t+1) )

~ , w 1, t must be bounded above.

Machine Learning: Perceptrons p.16/24

 Perceptron learning algorithm:
convergence
Lemma (worst case running time):
If the given problem is solvable, perceptron learning terminates after at most
(n + 1)2 2(n+1) log(n+1) iterations.
8e+07

7e+07

6e+07

5e+07

4e+07

3e+07

2e+07

1e+07

0
0 1 2 3 4 5 6 7 8

Exponential running time is a problem of the perceptron learning algorithm.

7
There are algorithms that solve the problem with complexity O(n 2 )

Machine Learning: Perceptrons p.17/24

 Perceptron learning algorithm:
cycle theorem
Lemma:
If a weight vector occurs twice during perceptron learning, the given task is
not solvable. (Remark: here, we mean with weight vector the extended
variant containing also w0 )
Proof: next slide

Machine Learning: Perceptrons p.18/24

 Perceptron learning algorithm:
cycle theorem
Lemma:
If a weight vector occurs twice during perceptron learning, the given task is
not solvable. (Remark: here, we mean with weight vector the extended
variant containing also w0 )
Proof: next slide
Lemma:
Starting the perceptron learning algorithm with weight vector ~
0 on an
unsolvable problem, at least one weight vector will occur twice.
Proof: omitted, see Minsky/Papert, Perceptrons

Machine Learning: Perceptrons p.18/24

 Perceptron learning algorithm:
cycle theorem
Proof:
~ (t+k)
Assume w =w ~ (t) . Meanwhile, the patterns ~x(t+1) , . . . , ~x(t+k) have been
applied. Without loss of generality, assume ~ x(t+1) , . . . , ~x(t+q) P and
~x(t+q+1) , . . . , ~x(t+k) N . Hence:
~ (t) = w
w ~ (t+k) = w
~ (t) + ~x(t+1) + + ~x(t+q) (~x(t+q+1) + + ~x(t+k) )
~x(t+1) + + ~x(t+q) = ~x(t+q+1) + + ~x(t+k)

Machine Learning: Perceptrons p.19/24

 Perceptron learning algorithm:
cycle theorem
Proof:
~ (t+k)
Assume w =w ~ (t) . Meanwhile, the patterns ~x(t+1) , . . . , ~x(t+k) have been
applied. Without loss of generality, assume ~ x(t+1) , . . . , ~x(t+q) P and
~x(t+q+1) , . . . , ~x(t+k) N . Hence:
~ (t) = w
w ~ (t+k) = w
~ (t) + ~x(t+1) + + ~x(t+q) (~x(t+q+1) + + ~x(t+k) )
~x(t+1) + + ~x(t+q) = ~x(t+q+1) + + ~x(t+k)
Assume, a solution w
~ exists. Then:
(
(t+i) 0 if i {1, . . . , q}
w
~ , ~x
<0 if i {q + 1, . . . , k}

Machine Learning: Perceptrons p.19/24

 Perceptron learning algorithm:
cycle theorem
Proof:
~ (t+k)
Assume w =w ~ (t) . Meanwhile, the patterns ~x(t+1) , . . . , ~x(t+k) have been
applied. Without loss of generality, assume ~ x(t+1) , . . . , ~x(t+q) P and
~x(t+q+1) , . . . , ~x(t+k) N . Hence:
~ (t) = w
w ~ (t+k) = w
~ (t) + ~x(t+1) + + ~x(t+q) (~x(t+q+1) + + ~x(t+k) )
~x(t+1) + + ~x(t+q) = ~x(t+q+1) + + ~x(t+k)
Assume, a solution w
~ exists. Then:
(
(t+i) 0 if i {1, . . . , q}
w
~ , ~x
<0 if i {q + 1, . . . , k}
Hence,
(t+1) (t+q)

w
~ , ~x + + ~x 0
(t+q+1) (t+k)

w
~ , ~x + + ~x <0 contradiction!
Machine Learning: Perceptrons p.19/24
 Perceptron learning algorithm:
Pocket algorithm
how can we determine a good
perceptron if the given task cannot
be solved perfectly?
good in the sense of: perceptron
makes minimal number of errors

Machine Learning: Perceptrons p.20/24

 Perceptron learning algorithm:
Pocket algorithm
how can we determine a good
perceptron if the given task cannot
be solved perfectly?
good in the sense of: perceptron
makes minimal number of errors

Machine Learning: Perceptrons p.20/24

 Perceptron learning algorithm:
Pocket algorithm
how can we determine a good Perceptron learning: the number of
perceptron if the given task cannot errors does not decrease
be solved perfectly? monotonically during learning
good in the sense of: perceptron Idea: memorise the best weight
makes minimal number of errors vector that has occured so far!
Pocket algorithm

Machine Learning: Perceptrons p.20/24

 Perceptron networks
perceptrons can only learn linearly
separable problems.
famous counterexample:
XOR(x1 , x2 ):
P = {(0, 1)T , (1, 0)T },
N = {(0, 0)T , (1, 1)T }

Machine Learning: Perceptrons p.21/24

 Perceptron networks
perceptrons can only learn linearly classifies three patterns
separable problems. accurately, e.g. w0 = 0.5,
famous counterexample: w1 = w2 = 1 classifies
XOR(x1 , x2 ): (0, 0)T , (0, 1)T , (1, 0)T but fails
P = {(0, 1)T , (1, 0)T }, on (1, 1)T
N = {(0, 0)T , (1, 1)T }
networks with several perceptrons
are computationally more powerful
(cf. McCullough/Pitts neurons)
lets try to find a network with two
perceptrons that can solve the XOR
problem:
first step: find a perceptron that

Machine Learning: Perceptrons p.21/24

 Perceptron networks
perceptrons can only learn linearly classifies three patterns
separable problems. accurately, e.g. w0 = 0.5,
famous counterexample: w1 = w2 = 1 classifies
XOR(x1 , x2 ): (0, 0)T , (0, 1)T , (1, 0)T but fails
P = {(0, 1)T , (1, 0)T }, on (1, 1)T
N = {(0, 0)T , (1, 1)T } second step: find a perceptron
that uses the output of the first
networks with several perceptrons
perceptron as additional input.
are computationally more powerful
Hence, training patterns are:
(cf. McCullough/Pitts neurons)
N = {(0, 0, 0), (1, 1, 1)},
lets try to find a network with two P = {(0, 1, 1), (1, 0, 1)}.
perceptrons that can solve the XOR perceptron learning yields:
problem: v0 = 1, v1 = v2 = 1,
first step: find a perceptron that v3 = 2

Machine Learning: Perceptrons p.21/24

 Perceptron networks
(cont.)
XOR-network:
1
y
x1 1
2 1
1

x2 1 1
0.5
1

Machine Learning: Perceptrons p.22/24

 Perceptron networks
(cont.)
XOR-network:
1
y
x1 1
2 1
1
2
x2 1 x2
1
0.5 1
+ -
1
0
Geometric interpretation: - +
x1
-1

-2
-2 -1 0 1 2

Machine Learning: Perceptrons p.22/24

 Perceptron networks
(cont.)
XOR-network:
1
y
x1 1
2 1
1
2
x2 1 x2
1
0.5 1
+ -
1
0
Geometric interpretation: - +
x1
partitioning of first perceptron -1

-2
-2 -1 0 1 2

Machine Learning: Perceptrons p.22/24

 Perceptron networks
(cont.)
XOR-network:
1
y
x1 1
2 1
1
2
x2 1 x2
1
0.5 1
+ -
1
0
Geometric interpretation: - +
x1
partitioning of second perceptron, assuming -1
first perceptron yields 0
-2
-2 -1 0 1 2

Machine Learning: Perceptrons p.22/24

 Perceptron networks
(cont.)
XOR-network:
1
y
x1 1
2 1
1
2
x2 1 x2
1
0.5 1
+ -
1
0
Geometric interpretation: - +
x1
partitioning of second perceptron, assuming -1
first perceptron yields 1
-2
-2 -1 0 1 2

Machine Learning: Perceptrons p.22/24

 Perceptron networks
(cont.)
XOR-network:
1
y
x1 1
2 1
1
2
x2 1 x2
1
0.5 1
+ -
1
0
Geometric interpretation: - +
x1
combining both -1

-2
-2 -1 0 1 2

Machine Learning: Perceptrons p.22/24

 Historical remarks
Rosenblatt perceptron (1958):
retinal input (array of pixels)
preprocessing level, calculation
of features
adaptive linear classifier
inspired by human vision

linear
retina features classifier

Machine Learning: Perceptrons p.23/24

 Historical remarks
Rosenblatt perceptron (1958): if features are complex enough,
retinal input (array of pixels) everything can be classified
preprocessing level, calculation if features are restricted (only
of features parts of the retinal pixels
adaptive linear classifier available to features), some
inspired by human vision interesting tasks cannot be
learned (Minsky/Papert, 1969)

linear
retina features classifier

Machine Learning: Perceptrons p.23/24

 Historical remarks
Rosenblatt perceptron (1958): if features are complex enough,
retinal input (array of pixels) everything can be classified
preprocessing level, calculation if features are restricted (only
of features parts of the retinal pixels
adaptive linear classifier available to features), some
inspired by human vision interesting tasks cannot be
learned (Minsky/Papert, 1969)
important idea: create features
instead of learning from raw data

linear
retina features classifier

Machine Learning: Perceptrons p.23/24

 Summary
Perceptrons are simple neurons with limited representation capabilites:
linear seperable functions only
simple but provably working learning algorithm
networks of perceptrons can overcome limitations
working in feature space may help to overcome limited representation
capability

Machine Learning: Perceptrons p.24/24