Follow along in the code from Steps I-VII, here
Here are a few exercises to further explore and understand the dl4j api.
-
Take a look at the csv files written out. Compare this to the raw data that can be viewed here
The format: one time series per file, and a separate file for the labels matched up by name.For example, train/features/0.csv is the features using with the labels file train/labels/0.csv The data here is a univariate time series, we only have one column in the CSV files. Furthermore, because we have only one label for each time series, the labels CSV files contain only a single value For more details on importing time series and how to handle time series with different lengths, refer here
-
Vary the mini batch size and note the effect on the score function in the UI. What happens when a larger mini batch size is used? A smaller?
-
Vary the learning rate by 100x,10x and 0.1x. Note the effect on the scores in the UI.
-
Leave out normalization. Does your net learn? Try another normalization technique.
-
Explore the different updaters and optimization algorithms.