Sharpened Cosine similarity
Sharpened Cosine similarity is a machine learning methodology initially designed to replace convolution in neural networks as a feature detector. It was first proposed by Brandon Rohrer[1] as a feature detector for the MNIST digits data set. Raphael Pisoni later wrote a blog post[2] on the first implementation as a tensorflow class.
History
The first mention of the method was on 24 February 2022 by Brandon Rohrer (iRobot, formerly at Facebook, Microsoft, DuPont Pioneer, Sandia Labs, MIT).
Since then, there have been several contributions by other individuals listed below:
2022-02-18 blog post by Raphael Pisoni. SOTA parameter efficiency on MNIST. Intuitive feature interpretation.
2022-02-17 PyTorch code by Brandon Rohrer. SCS model with 95.3k parameters and 15.9% error on CIFAR-10.
2022-02-16 PyTorch code by Brandon. SCS model with 68k parameters and 18.4% error on CIFAR-10.
2022-02-14 PyTorch code by Brandon. PyTorch implementation of SCS running on Fashion MNIST.
2022-02-01 PyTorch code by Stephen Hogg. PyTorch implementation of SCS.
2022-02-01 PyTorch code by Oliver Batchelor. PyTorch implementation of SCS.
2022-01-31 PyTorch code by Ze Wang. PyTorch implementation of SCS.
2022-01-30 Keras code by Brandon. Keras implementation of SCS running on Fashion MNIST.
2022-01-06 blog post by Raphael. Description of SCS. Announcement of Keras implementation.
2022-01-06 Keras code by Raphael. Keras implementation of SCS.
2020-02-24. Twitter thread by Brandon.[1] Justification and introduction of SCS.
References
- ↑ 1.0 1.1 "https://twitter.com/_brohrer_/status/1232063619657093120". Twitter. Retrieved 2022-02-20. External link in
|title=(help) - ↑ Pisoni, Raphael (2022-01-06). "Sharpened Cosine Distance as an Alternative for Convolutions". rpisoni.dev. Retrieved 2022-02-20.
This article "Sharpened Cosine similarity" is from Wikipedia. The list of its authors can be seen in its historical and/or the page Edithistory:Sharpened Cosine similarity. Articles copied from Draft Namespace on Wikipedia could be seen on the Draft Namespace of Wikipedia and not main one.
