I am working with a heavily imbalanced dataset and attempted to resample the data to get a more even balance. No matter what I do my classifier seems to produce the same valued predictions no matter what input is passed to the network. Here is my current Jupyter notebook where I am seeing the issue: https://gist.github.com/nickrobinson/465919453774c0b4f7fb36055f030852. Is this an artifact of my imbalanced dataset, or am I doing something wrong with the Gluon API?
it’s possible that the entries in your test data is just the same feature vector repeated multiple times. You should print out test_data[:100] to see if that’s the case. That is if you haven’t already resolved this.