![](https://ccweb.imgix.net/https%3A%2F%2Fwww.classcentral.com%2Fimages%2Ficon-black-friday.png?auto=format&ixlib=php-4.1.0&s=fe56b83c82babb2f8fce47a2aed2f85d)
Overview
![](https://ccweb.imgix.net/https%3A%2F%2Fwww.classcentral.com%2Fimages%2Ficon-black-friday.png?auto=format&ixlib=php-4.1.0&s=fe56b83c82babb2f8fce47a2aed2f85d)
This course teaches about a novel neural network architecture called SpineNet, which challenges traditional CNN design by incorporating scale-permuted intermediate features and long-range skip connections. The learning outcomes include understanding the limitations of current CNN architectures, exploring the concept of scale-permuted networks, and learning about the benefits of SpineNet over traditional models. The course covers topics such as neural architecture search, up- and downsampling techniques, and the transition from ResNet to SpineNet. The intended audience for this course is individuals interested in machine learning, artificial intelligence, and neural network design. The teaching method involves a detailed explanation of the paper, including key concepts, experiments, and conclusions.
Syllabus
- Intro & Overview
- Problem Statement
- The Problem with Current Architectures
- Scale-Permuted Networks
- Neural Architecture Search
- Up- and Downsampling
- From ResNet to SpineNet
- Ablations
- My Idea: Attention Routing for CNNs
- More Experiments
- Conclusion & Comments
Taught by
Yannic Kilcher