1.   Abstract ideas became focused as he pulled together previous work on neural nets.

2.   Any learning neural net explores a space in which each state is described by a large set of simple parameters.

3.   For neural nets and genetic algorithms, it is not so much fallible as crude.

4.   I am grateful to Teresa Ludermir who introduced me to it, and to logical neural nets.

5.   In all three instances, neural nets were being used to make routine Decisions that otherwise would have required human intervention.

6.   Neural nets can learn rule structures and patterns on their own from historical data or through experience.

7.   Training sets are typically large, for any kind of neural net.

8.   Unlike less sophisticated Al, neural nets do not require that elaborate rule structures be specified in advance.

9.   This theory has also been phrased in terms of firing thresholds in neural nets.

10.   Given a stream of data, a neural net can learn to identify subtle patterns.