From the course: Deep Learning: Model Optimization and Tuning
Join today to access over 24,900 courses taught by industry experts.
Regularization experiment
From the course: Deep Learning: Model Optimization and Tuning
Regularization experiment
- [Instructor] The code for experiments in this chapter are available in the notebook code_04_XX Overfitting Management. The regularization experiment is in section 4.3. Here we are going to compare three options, L1, L2, and L1L2. We past them as options when creating the hidden layers. Note that we are comparing validation accuracy for the overfitting experiments. Let's run the experiment and review the results. In the plot we see that L2 and L1L2 seem to perform better than L1. For the Iris dataset the validation dataset is very small. So repeated runs may not provide consistent results. In real world experiments, it is recommended to have a larger validation set to achieve consistent results across multiple runs.
Download courses and learn on the go
Watch courses on your mobile device without an internet connection. Download courses using your iOS or Android LinkedIn Learning app.