using 2D data as input for a neural network

have data that i'd like to use as input for a pattern recognition neural net project

the data comes as a series of tables, each of which needs to be input all at once, rather than one row or one col at a time

so, I figure it needs to be "spooled" out to one long array, and the net will need at least that many input neurons

so, two questions:

1) does it matter how it gets spooled out? say, column by column, row by row, or either as long as I'm consistent?

2) the tables are different sizes, should I just pad the smaller ones out with zeros? nulls?

thanks

Answer 1

[Greg Heath]

0. An input layer node is not a neuron. Only hidden and output layer nodes are neurons.

1. Use the colon operator matrix(:) to convert matrices to a single column.

2. No need to pad with zeros or nulls. Just stack the converted matrices ([a(:);b(:);...;z(:)]

Hope this helps.

Thank you for formally accepting my answer

Greg

[59morgan]

thx