Check out the follow-up video:
How to Design a Neural Network | 2020 Edition
• How to Design a Neural Network | 2020...
Designing a good model usually involves a lot of trial and error. It is still more of an art than science. The tricks and design patterns that I present in this video are mostly based on 'folk wisdom', my personal experience, and ideas that come from successful model architectures.
Deep Learning Crash Course playlist: • Deep Learning Crash Course
Previous video:
• Convolutional Neural Networks Explained
Highlights:
How to choose the number of layers
Deeper vs wider models
Design patterns and hyperparameters
Skip connections
ResNet
Inception module
Fully Convolutional Networks
Pointwise convolutions
Dimensionality reduction
MobileNets
Separable convolutions
How to choose sliding window stride
How to choose pooling parameters
How to choose activation function
What type of regularization to use
How to choose the batch size
Further reading:
Deep Learning by Ian Goodfellow:
http://www.deeplearningbook.org/
CS231n: Convolutional Neural Networks for Visual Recognition
http://cs231n.github.io/
Deep Residual Learning for Image Recognition
https://arxiv.org/pdf/1512.03385.pdf
Fully convolutional networks for semantic segmentation
https://www.cv-foundation.org/openacc...
Surface Water Mapping by Deep Learning
http://www.isikdogan.com/files/isikdo...
Network In Network
https://arxiv.org/pdf/1312.4400.pdf
Rethinking the inception architecture for computer vision
https://www.cv-foundation.org/openacc...
Mobilenets: Efficient convolutional neural networks for mobile vision applications
https://arxiv.org/pdf/1704.04861.pdf
