Skip to content

Details

# Interactive Zoom Meetup #

This will be a live presentation over Zoom complete with Q&A.

We recommend you connect to the meeting 5 minutes early to ensure you can connect successfully and test your audio in Zoom.

https://us02web.zoom.us/j/88622548112?pwd=a0lvWTdhL0M4emxJNHdCQW5SbmkxQT09

If you'd like to participate in the Q&A, please sign up to Discord ahead of time. A link to join our Discord chat room will be provide on the night.

==================

Embed all the things: Neural Embeddings for Categorical Data
William Caicedo

It's often said that Neural Networks should not be the first option to consider when working with tabular data, as there are other techniques that offer better performance (i.e. gradient boosting). However, an often overlooked strength of Neural Networks is their ability to learn useful vector representations that can be used in downstream tasks.

For example, Word2Vec learns a transformation that generates dense vectors - called embeddings - that capture semantic and contextual information. These embeddings can be used for NLP tasks such as sentiment analysis. It turns out that the same principle can be applied to tabular/categorical data in general to obtain semantically meaningful representations that can then be used for prediction purposes.

In this talk, William will discuss the basics of using Neural Networks to create embeddings for categorical and tabular data, and some interesting bits about the application of neural embeddings to the development of recommender systems.

==================

William Caicedo is senior data scientist at Movio, and a PhD candidate at the AUT School of Engineering, Computer and Mathematical Sciences. His work at Movio involves the development of recommender systems using Deep Learning.

William received a BSc in Computer Science and a MSc in Engineering from Universidad Tecnológica de Bolívar (Colombia). He was awarded a PhD scholarship from AUT’s School of Engineering, Computer and Mathematical Sciences in 2017.

==================

Our Key Sponsor is IBM

We'd like to thank David Stokes (https://www.linkedin.com/in/davidstokes/) for volunteering his time and expertise in facilitating the conference online platform.

Thanks also to ProductTank Auckland (https://www.meetup.com/ProductTank-Auckland/) for leading an initiative to support us and other NZ meetups to continue adding value to their members during these difficult times.

Our meetups are inclusive and harassment is not tolerated. For more information, see our Code of Conduct: https://www.meetup.com/Auckland-AI-Meetup/messages/boards/thread/50720238

You may also like