Diagnosing and addressing the bias-variance tradeoff

You train a supervised model and observe the following: training-set error is 2%, while validation-set error is 18%. The Bayes (irreducible) error for this task is estimated to be around 1%.

(a) Diagnose what is happening in terms of bias and variance, and justify your diagnosis using the numbers given. (b) Propose at least three concrete interventions you would try, and for each, explain the mechanism by which it should help. (c) Explain how getting more training data would (or would not) help here, and contrast that with a different scenario where the train and validation errors were both around 17%.