This is a tiny data set of the first 50 records (10 random hyper-parameter combinations at 5 epochs) from the complete dataset of 10,000 rows) and for demonstration purpose only.
file | description |
---|---|
input.txt | input data generated with xor shape at noise 25%. seven features (X1, X2, X12, X22, X1X2, sin(X1), sin(X2)) and labels (ground truth) for 200 input points. The ratio of the training to test is a varying hyper-parameter |
index.txt | values for the hyper-parameters used for each run and summary statistics. There are 5 rows per parameter combination correponding to 5 different epochs (25, 50, 100, 200, 400) |
paramInfo.txt | information about each column of index.txt (hyper-parameters + output statistics) |
runs/[i].txt | one file for each run: predicted labels the points at the i'th run |
images/[i].png | one plot for each run output: predicted classification. training and test points are represented with white and black stroke respectively. |
The following images are the classification output plot for random hyper-parameters at different epochs.
epoch: 25 | epoch: 50 | epoch: 100 | epoch: 200 | epoch: 400 |
---|---|---|---|---|