This notebook presents regularization and supplemental approaches like initialization that are proven to improve the predictive accuracy of models in Deep Learning.The AI Soccer Coach's accuracy improves to 95%!
What to use this notebook for:
- Access regularization techiques such as L1, L2 and dropout to improve predictive accuracy.
- Develop intuition for how regularization works in order to apply it successfully.
- Supplement regularization with the appropriate choice of method for weight initiatialization.
Deep Learning models derive power from a large number of tunable parameters that impart capacity for learning. However, that can turn out to be too much of a good thing when making suitable choices for hyper-paramters that influence training. The approaches presented here help to guide wise choices.