node = neuron

n = neurons

m = layers

n*m words of storage

Architecture

• feed forward (instinctive)
• feed back (learning)

Training

• supervisor (pattern)
• unsupervised

Linear

• one fact per neuron

Non-linear

• many to many connections – back propagation

• Constructed (single layer)
• Trained

1. Define problem in terms of neural layers (0/0)
2. Represent information (select data type and locate data(historical data evaluation))

Patterns: catagorizers, recognizers, associators

3. Define network
4. Train network
5. Test Network

Structuring Data

1. Use randomly ordered facts
2. Show representation data
3. Neurons can’t be coded

1 = horse #1

2 = horse #2

4. Historical Data
5. Differences better than big numbers
6. Seasonal data
7. Like a lot of inputs and outputs
8. Think Qualitatively