digantamisra98/Mish
Research repository for Mish, a self-regularized non-monotonic neural activation function for deep learning.
Not currently ranked — collecting fresh signals.
star history
This repository contains the official implementation and experiments for the Mish activation function published at BMVC 2020. Mish is a non-monotonic neural activation function designed as a self-regularized alternative to functions like ReLU and Swish. The repository provides Jupyter notebooks demonstrating Mish’s effectiveness in computer vision tasks including image classification on CIFAR and ImageNet, as well as object detection benchmarks.
Frequently asked
- What is digantamisra98/Mish?
- Research repository for Mish, a self-regularized non-monotonic neural activation function for deep learning.
- Is Mish open source?
- Yes — digantamisra98/Mish is open source, released under the MIT license.
- What language is Mish written in?
- digantamisra98/Mish is primarily written in Jupyter Notebook.
- How popular is Mish?
- digantamisra98/Mish has 1.3k stars on GitHub.
- Where can I find Mish?
- digantamisra98/Mish is on GitHub at https://github.com/digantamisra98/Mish.