Variable Selection Lasso
Conditional variable screening is a popular method. It is used before variable selection methods such as LASSO to narrow down a few important
variables. Conditional variable screening further improves the method since it helps to reduce false positives and negatives. Apply machine learning method for high-dimensional models How to conduct conditional variable screening for deep Neural Network (NN)?
Construct test via developing test statistics for derivative estimators
Two key questions:
1 How to construct a high-quality derivative estimator using deep NN?
2 How to construct the test statistics?