Optimize for Recall or Precision

Course Q&A Deep Learning Specialization Improving Deep Neural Networks: Hyperparameter tun
1

Hello

This course mainly uses Accuracy to evaluate the Neural Net and discusses techniques to improve accuracy.

In some instances, I may want to optimize my NN for a high recall (within some precision limit). How would I go about that ? What tools could I use to differentiate recall and precision in my optimization process ?

Thanks

2

Hey @Leonard_Bouygues1,

Well after making sure that you understand precision and recall which i think you already know both you can follow these steps:

  • Set your precision limit “minimum acceptable precision level you require; this will be your constraint.”

  • Choose the Right Metric for example F1-Score or the Precision-Recall Curve.

  • Adjust the Threshold : choose a threshold that meets your desired precision limit.

  • Hyperparameter Tuning : Tune neural network hyperparameters (e.g., learning rate, layers, batch size) to optimize for your recall goal.

I hope it helps you.
Regards,
Jamal