hi,
I was wondering how we tune hyperparameters in unsupervised learning (for ex k in K-means). Since we don't really have a train-validation-test set split.
thanks
Lab 11, exercise 2 exactly deals with your question: https://github.com/epfml/ML_course/blob/master/labs/ex11/exercise11.pdf
For example, looking where the (training) cost function decreases rapidly before plateauing is one of the ways to choose \(k\). See e.g. https://medium.com/analytics-vidhya/how-to-determine-the-optimal-k-for-k-means-708505d204eb
unsupervised learning
hi,
I was wondering how we tune hyperparameters in unsupervised learning (for ex k in K-means). Since we don't really have a train-validation-test set split.
thanks
1
Lab 11, exercise 2 exactly deals with your question:
https://github.com/epfml/ML_course/blob/master/labs/ex11/exercise11.pdf
For example, looking where the (training) cost function decreases rapidly before plateauing is one of the ways to choose \(k\).
See e.g. https://medium.com/analytics-vidhya/how-to-determine-the-optimal-k-for-k-means-708505d204eb
1
Add comment