Performance of Neural Networks

1
Performance of Neural Networks
2
Materials Science
Brief explanation of neural networks
Four classes of innovations based on neural
networks
Are papers on neural networks respectable?
3
Empirical Equations
4
(No Transcript)
5
(No Transcript)
6
noise
7
modelling uncertainty
8
elegantly answers the question as to whether
sufficient data have been used to create the model
unreasonable to ask if sufficient data used to
create model, relative to number of
coefficients - may be sufficient data in some
regions of input space and not in others
9
International Fusion Reactor
Reduced activation steels
10
Journal of Nuclear Materials 348 (2006)
311-328 Kemp, Cottrell Bhadeshia
11
Exploitation of neural networks

• Discover new science
• Explain observations
• Design materials or processes
• Quantitative expression of data

12
Neural networks unexpected outcomes
13
(No Transcript)
14
(No Transcript)
15
(No Transcript)
16
(No Transcript)
17
tested at room temperature
18
tested at 100 C
19
(No Transcript)
20
Neural networks design
21
(No Transcript)
22
7Ni 2Mn
Keehan, Karlsson, Andrén, Bhadeshia, Science
Techn. Welding Joining 11 (2006) 9-18
23
Exploitation of neural networks

• Discover new science
• Explain observations
• Design materials or processes
• Quantitative expression of data

24
(No Transcript)
25
9Cr1Mo
Dimitriu Bhadeshia, 2007
26
Dimitriu Bhadeshia, 2007
27
Components of Creep Strength 2.25Cr1Mo
iron microstructure
550 C
solid solution
600 C
precipitates
Murugananth Bhadeshia, 2001
28
(No Transcript)
29
Exploitation of neural networks

• Discover new science
• Explain observations
• Design materials, processes,
• experiments
• Quantitative expression of data

30
Suppose we fail to achieve 650C Ferritic
Creep-Resistant Steel
31
(No Transcript)
32
(No Transcript)
33
Exploitation of neural networks

• Discover new science
• Explain observations
• Design materials, processes,
• experiments
• Quantitative expression of data

34
weld pool shape Mishra and DebRoy, MSE A, 454
(2007) 454
35
(No Transcript)
36
Five steps in the creation of meaningful neural
networks
37
(No Transcript)
38
Creation of model 1
Make predictions using model 2
Investigate prediction experimentally 2
Modelling uncertainty 2
Model or data disseminated 3
39
(No Transcript)