ML Researcher and Engineer with one leg in industry and another in academia
Bern
Joined November 2022
- Tweets5
- Following42
- Followers13
- Likes11
I made a machine learning puzzle: Find out what‘s going on in the different models and data sets to free the ghost of the latent space.
Check it out here for some fun during the holiday break:
github.com/mattminder/latent…
Matthias Minder retweeted
Great blog post! I think the experiments confirm my speculation that information distance works well with compression algorithms that take advantage of overlapped substrings information. Maybe even LCS+knn can work? hash n-gram algorithm is an awesome idea, love it.
@ZhiyingJ
Digging a bit deeper into the "GZIP beats BERT" paper, I think that a large part of why it works is because it compares character n-grams between documents. You can use this to make the implementation O(n) instead of O(n^2).
Here's a write-up: towardsdatascience.com/is-it…
