Does tensorflow have a way of calculating input importance for simple neural networks

16 views Asked by At

I have a small neural network in Tensorflow, with about 100 input variables and 2 hidden layers. Does Tensorflow have a built-in way of assessing input variable importance?

My first thought is to just drop each variable and calculate the accuracy, and the bigger the drop in accuracy, the more important the variable. However, there is randomness inherent in Tensorflow (I guess?) so every trial is a little different and it makes it difficult.

0

There are 0 answers