Purpose

To create a neural net that can read images of numbers and output what they are.

Efficiency

Must have as fast an run time as possible, this is so that future adaptations might be possible, which require growing the net such that there are more operations needed per iteration.

UI

There should be a user interface so that a user can teach the net new things, verify that it has indeed learnt training regimes, etc. The UI should be simple with not too many fields, yet still retain high functionality.

Future

Must be adaptable to future possible implementations such as reading letters, and perhaps having the brain interact with the environment.


class diagram