News

Free Tech Talk: Stellar classification using machine learning

CCA_barclays_lab3

Twinkle twinkle little star, how I wonder what you are…
This free evening talk will explore Machine Learning applications in astrophysics, using manual star classification techniques as an example.

Abstract

In my previous life, I was an astronomer and one of the big tasks many PhD students face is manual star classification. But why ask a student to do what a machine should be able to do too? In this talk, I use the spectra (stellar flux vs wavelength) of identified stars to build a classifier which will detect the identity of the stars. Using a very simple non linear SVM, I achieve an 86% accuracy with my model. The next step is to use deep neural nets to achieve a better accuracy.

Speaker: Sandra Greiss

I joined Lyst in December 2014, as a junior data scientist, a couple of months after I finished my PhD in Astronomy and Astrophysics (Warwick). Lyst is an online fashion website, where we aggregate millions of products from thousands of retailers. I am responsible for building the ‘autocomplete’ and ‘did you mean’ features of Search on our website. I have now moved towards more data science projects, where I want to predict the fabric of clothes, shoes and bags on Lyst using their descriptions and images. I use crowdsourcing to tag the products for the training dataset of my models. Besides work, I have a weird obsession with shoes, good food and travelling every few weeks! http://www.lyst.com

If you’re curious about Machine Learning and the latest innovations in tech, we hope to see you there. Anyone welcome!

Please register for the event below:

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s