From the course: PyTorch Essential Training: Working with Images
Join today to access over 24,400 courses taught by industry experts.
Understanding convolutional neural networks (CNNs)
From the course: PyTorch Essential Training: Working with Images
Understanding convolutional neural networks (CNNs)
- [Instructor] The biggest challenge when working with neural networks is overfitting. The more powerful a neural network is, meaning the more layers and weights it has, the more it's prone to overfitting. It usually happens when a neural network learns the noise in the dataset instead of making decisions based on the true signal. Wait, what does it mean? Let's take a look at these two images. Obviously, both images represent a cat. In the first picture, the cat is sitting on a blanket on a nice chair in the room. In the second picture, we have a silhouette of a cat. So everything surrounding the cat in the first picture, blanket, chair, background, is the noise. In the second picture, the middle black part of the cat is also noise. It doesn't describe a cat, it just fills the middle. A signal in these two pictures is the essence of the cat. Overfitting also happens when we have more parameters than necessary to learn a specific dataset and not enough training examples. We would like…
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.
