https://github.com/heardacat/ml-speedrun
speedrunning machine learning
https://github.com/heardacat/ml-speedrun
Last synced: 2 months ago
JSON representation
speedrunning machine learning
- Host: GitHub
- URL: https://github.com/heardacat/ml-speedrun
- Owner: HeardACat
- Created: 2020-12-19T11:14:40.000Z (over 4 years ago)
- Default Branch: main
- Last Pushed: 2020-12-19T22:42:46.000Z (over 4 years ago)
- Last Synced: 2025-03-26T16:40:29.724Z (2 months ago)
- Size: 4.88 KB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Machine Learning Speedrun
Speedrunning Machine Learning
In this speedrun I'll attempt to re-implement a number of algorithms which define the "bread-and-butter" of what a data scientist should know (maybe). I will use vanilla "numpy" to implement the following...
* linear models (logistic regression?)
* naive bayes
* decision trees
* boosting/stacking of some variation
* convolution networks
* some kind of hacked version of u-net for image segmentation
* soft actor critic
I won't be using any frameworks (besides reading in data), and will probably have to implement some framework (rather badly) from scratch - so we'll see how we go.
The question really will be how far will I get?
-----
The scoring system will be based on some benchmark datasets.
* We'll use iris dataset, and newsgroup to start off with for the easier algorithms
* (If I get there) the image data will be based on Pascal VOC 2005: http://host.robots.ox.ac.uk/pascal/VOC/voc2005/index.html, which will support both classification and image segmentation tasks