Erudite CSV Files - adam-nnl/Erudite GitHub Wiki

Erudite accepts two types of comma separated value(.csv) files, one for testing data sets and another slightly different version for training testing sets.

Training data .csv

Training data .csv files are row based. The first row in the file should have comma separated values for-number_of_training_set_examples,inputs,outputs

The following row should be the comma seperated input values for the first training pattern. Followed by a row of comma separated values for the expected/desired output values.

The following example is a four example training set for a simple XOR network that uses a sigmoid[0.0 - 1.0] function:

4,2,1
0,1
1
0,0
0
1,0
1
1,1
0

---

Testing data .csv

Testing data set .csv files are also row based.

The first row should have comma separated values for- #of-test-examples,inputs

After the first row there should be row pairs for each of the test examples. The first row in the pair should be a string denoting the example name. The second row for each example pair is the comma seperated test input values.

As an example this would be a test set csv to use on a XOR neural network:

4,2
example1-0
1,0
example0-0
0,0
example0-1
0,1
example1-1
1,1