What is the ratio of training dataset, test dataset and validation dataset?

[tennybritz]

The experiment worked out pretty well. A good job.

[navfour]

Thanks，in“config_function.py” ,10% of the data set as the test dataset.

[tennybritz]

So there is no validation dataset. Is it because the amount of data is too small?

[navfour]

Sorry, I made a slip of the pen yesterday。In fact, there is no test dataset in this code，the 10% is the percentage of val data. But it can set the test dataset by Additional experiments. For example, we tried to put one year of GDP data points (4 points) out of the dataset. After moedl training , we compare the predictions of this 4 data points ，this 4 piont is the size of test dataset，it can also use other proportions.