Ensemble Clustering Evaluation Issues.
EClust are often evaluated using NMI and ARI. If they were good measure, you would get a score of one for a perfect consensus, and 0 for a very bad one. There are very simple experiments showing they do not behave this way. This article will show the problem and describe where it comes from. Next, we will propose alternative measure this consensus functions.
Tags: clustering ensemble machine learning scoring unsupervised
Ensemble Clustering.
Ensemble clustering aims to combine multiple clustering together to get a better consensus clustering. Here, I will present the main approaches, and how evaluation is performed. I will introduce some reflection about scoring and diversity.
Tags: machine learning unsupervised clustering ensemble
Clustering algorithms
A short introduction to clustering algorithms and their related problems. Disclaimer - this is not an introduction to KMeans or DBSCAN, here, the main concept and evaluation strategies are presented.
Tags: machine learning clustering black box understanding
Entropy
Tags:
Distance, Divergence, and Similarity - A Cheat Sheet
When you read around thousand of papers a year, you become aware that there are many ways to measure distance or similarity between stuff. They do not apply to the same object, the same data-structure, so this variety is necessary.
Tags: distance metric loss similarity divergence scoring
Writing is not that easy
Writing is hard. It is not as easy as pressing the "enter" button to publish your post. It requires more than writing down a bunch of ideas. People cannot read in our mind, therefore we need to organize ideas in a coherent way, around a red line or a story. Additionally, finishing everything requires commitment, checking everything, arranging stuff in a nice way. Nevertheless, even if it is really time consuming, you learn plenty of stuff.
Tags: tech writting blog being famous
See the full post list
Information about pagination