computer science - Neural Network to identify Seven-Segment Numerals -

i studying machine learning , working on first neural network project 1 of classes. programming network in java. point of network identify seven-segmented numeral (like on regular digital clock). network not have linked real sensors, needs work in theory based on inputs 0's , 1's in text form, not binary, correspond hypothetical sensor matrix laid across top of number.

my question is, sort of output looking get?

1. will binary output correspond same sort of matrix input or binary output supposed represent input number in binary such returning 111 7?
2. if return matrix, point of network?

the input seven-segment numeral (1 x 7) vector, 1 segments on , 0 segments off.

as output, don't specify want be, let's assume want tell "which digit screen showing". since there 10 digits (0 through 9), have 10 possible answers. output (1 x 10) vector, each number corresponding 1 of digits. value represents how confident network correct answer (typically output values lie in [0, 1], depends on setup) ideally want network return vector having 1 on 1 attribute , zeros in others.

note however, case classifier not useful. classification algorithm generalizes has seen in past. so, useful handwritten recognition, because if same person writes same digit twice, not same. in case, each digit same across 7-segment displays, network not learning, rather memorizing input.