Computers - Using your Brains

1
Computers - Using your Brains

• Jim Austin
• Professor of Neural Computation

2
(No Transcript)
3
(No Transcript)
4
Pentium III
5

• So how complex is it ?
• 1012 neurons 1,000,000,000,000
• 1000 connections between neurons.
• One brain can hold ...
• 1,000,000,000,000,000 numbers!

6
What do 1012 neurons look like ?

• 1600 times Population of the world
  (6,100,000,000)
• 78,125 times the complexity of the Pentium III
• Equal to the number of stars in our galaxy

4 Meters
4 Meters
4 Meters
Sand
7
The good and the bad
8
Why are computers so restricted ?
ACE
9
Leo - for stock control
10
Colossus - for breaking codes
11
Pegusus - for scientific work
12
Neurons verses Gates
Input 1
Output
Input 2
NAND Gate
Boolean Logic - both inputs OK, output not OK
13
Gates - NAND
ALL inputs to be OK for output to be NOT OK
Output
Input 1
Input 2

14
(No Transcript)
15
Evolution ?
Should have picked a NAND gate for the brain...
16
Neuron
Output threshold (input A x weight A input B
x weight B)
A

Inputs
Output
B
Weights
Threshold logic - threshold 1 - one or more
inputs OK ? output OK
17
Neuron
At least one OK for output to be OK

At least three OKs for output to be OK
18
Weights

• Can also alter connections/importance of inputs
• using the weights on the inputs

1
0
1
1

3.5
0.5
1
1
19
Why did this difference develop ?

• The analysis of the operation of a machine using
  two indication elements and signals can be
  conveniently be expressed in terms of a
  diagrammatic notation introduced, in this
  context, by Von Neuman and extended by Turing.
  This was adopted from a notation used by Pits and
  McCulloch as a possible way of analyzing the
  operation of the nervous system, Calculating
  Instruments Machines, D Hartree, 1950,
  Cambridge University Press.
• Probably dropped due to the development of the
  silicon chip
• simpler to build Boolean logic gates rather than
  neuron units.

20
Functional elements.
Threshold n gate k ?n
1
z
Excitation, OR
2
z
Excitation, AND
21
ICT Orion Computer

• Used Neuron logic - 1962

22
(No Transcript)
23
Learning !

• Learning at neuron level
• Adjustment of which inputs are important
• Conventional computers have no implicit learning
  ability

24
Spot the difference
25
(No Transcript)
26
(No Transcript)
27
(No Transcript)
28
(No Transcript)
29
Can we build useful systems with neurons ?

• Better tolerance to failure
• Parallelism/use of threshold logic/distributed
  memory
• Faster operation
• Massive parallelism
• Better access to uncertain information
• Threshold logic/neurons
• Where the inputs are uncertain
• Threshold logic/neurons.
• Where we want low power
• Asynchronous systems
• Adaptability
• Use of weights and learning methods.

30
So what have we done with these ?

• Cortex-1
• 28 Processor cards, each holding 128 hardware
  neurons.
• Each with 1,000,000,000 weights.
• 16MHz.
• PCI based card.

31
Complete Machine 400,000,000 neuron
evaluations per second 28,000 inputs 30 bits set
on input 1,000,000 neurons.
32
Cortex-1 node
5,120,000,000 neuron weights, 640 neurons.
33
Recognising Addresses for the Post Office
34
Recognising trademarks
35
Text search engines

• Tolerant to spelling errors.
• Finds similar words to those supplied, for
  example chair, seat, bench.
• Learns these similarities automatically from
  text.
• Uses neural engine for document storage.
• Estimated 400,000,000 documents searched per
  second.

36
Molecular Databases

• One of few systems that deal with the full 3D
  molecule

37
(No Transcript)
38
Query
Good matches
Bad Match
39

Thanks...
Aaron Turner Mick Turner Vicky Hodge Julian
Young Anthony Moulds Zyg Ulanowski Ken
Lees Michael Weeks Sujeewa Alwis John
Kennedy David Lomas and many others .
(Its Brains from Thunderbirds !)