Artificial Intelligence By Example

Scheduling and blockchains

An artificial neural network (ANN) system built on dimension reduction might sometimes be seriously underfitted. As long as it was trained with ready-to-use datasets (MNIST, CIFAR, and others), the system worked fine. It even ran perfectly on small datasets.

Then all of a sudden, on the first week in the food processing industry, for example, it mistakes a positive gap (an available resource) for a negative gap (not enough production).

For example, since a gap meant missing products for several days, the AI system wrongly predicted a negative gap when it was not supposed to. In this circumstance, this was a deliberate choice of the management to slow production down due to resource shortages (see Chapter 12, Automated ...