Non-Sequential Data

Non-Sequential Data Type. (Advanced users only)

A non-sequential raw data type is data where you have your data prearranged by sample sets into a spreadsheet. This data can then be exported to tab delimited text by your spreadsheet editor and used by Psychonet as predefined sample sets.

This data type is provided for special case needs where output is a computed value.

Warning: This may sound straight forward and perhaps easier. You're using a spreadsheet right? Why is this an advanced setting?

The answer as follows: When you're working with spreadsheet data you need to know in advance what your neural network configuration will be. The key question being "How many inputs/outputs will you have?" This is an extremely difficult question and there is no exact formula to determine an answer for you.

If would like to explore the usage of this data type further we would suggest you first use a sequential (streamed) data example. Look at the training data that is generated. Can you make a spreadsheet like that?

If all else fails there is also an example of spreadsheet data file included in the examples directory. This might be useful to look at as well.