Day 6 - Unsupervised Learning

Posted on June 5, 2017 by Govind Gopakumar

Lecture slides in pdf form

Prelude

Announcements

Story so far

Supervised Learning

Techniques

Clustering

Clustering - I

Why do we need it?

What’s the easiest way to do it?

Clustering - II

Model overview

Training the model

Clustering - III

Model parameters

How to find both?

Clustering - IV

Alternating optimization

Finding the parameters

Clustering - V

Estimating the cluster IDs

Estimating the cluster means

Clustering - VI

Geometry of the model

Uniqueness of clustering

Clustering - VII

Comments about K-Means

Limitations

Smarter Clustering

Gaussian Mixture Models - I

Why should we improve our clustering?

Generative modelling

Gaussian Mixture Models - II

Review of the Gaussian distribution

Estimation of a Gaussian

Gaussian Mixture Models - III

Modelling assumptions

Model overview

Gaussian Mixture Models - IV

Alternating optimization?

What are the parameters then?

Conclusion

Concluding Remarks

Takeaways

Announcements

References