From the course: Computer Vision for Data Scientists
Join today to access over 24,900 courses taught by industry experts.
Regularization and data augmentation
From the course: Computer Vision for Data Scientists
Regularization and data augmentation
- [Instructor] Regularization and data augmentation are essential for training convolutional neural networks that achieve good generalization and don't overfit. Remember that overfitting happens when your model performs really well on your training data, but doesn't perform well on your testing data. That means that it's not generalizing to new and unseen data. There are a lot of different regularization techniques and data augmentation strategies out there that you can use to address this issue. Let's go ahead and talk about a few of them. So first, let's define what regularization even is. Regularization methods add a penalty term to the loss function, and this encourages the model to learn simpler and more general patterns in the data. Dropout is a popular regularization technique in CNNs. Now, this technique doesn't adjust the loss function at all but what it does is it randomly drops out neurons during training. So at…
Practice while you learn with exercise files
Download the files the instructor uses to teach the course. Follow along and learn by watching, listening and practicing.
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.