[Pycon] [new paper] "Pietro Mascolo" - A primer on manifold dimensionality reduction techniques

[info a pycon.it]

Title: A primer on manifold dimensionality reduction techniques
Duration: 45 (includes Q&A)
Q&A Session: 0
Language: en
Type: Talk

Abstract: When dealing with large datasets in machine learning problems, we often face issues with noisy, collinear, or irrelevant features.
Dimensionality reduction techniques help ameliorate these issues by finding a better representation of the data that maintains explainability while being simpler to understand and to use.

Dimensionality reduction is the process of reducing the number of random variables under consideration, by obtaining a new set of principal variables. While some simpler techniques for feature selection and feature extraction are commonly used and understood (e.g. PCA and all its variants), these techniques lack the power of some more advanced manifold techniques such as Isomal, LLE, T-SNE, and UMAP.

In this talk we will explore some manifold learning algorithms to achieve feature extraction from a complex dataset; we will try to understand how the algorithms work under the hood and we will try to understand how to extract maximum value from the results.



Tags: [u'features', u'Machine Learning', u'dimensionality-reduction']