Wednesday, June 26, 2019 at 6:00:00 PM 1740 Main St 1740 Main St, Kansas City, MO

The Human Protein Atlas Kaggle competition (https://www.kaggle.com/c/human-protein-atlas-image-classification) presented a multi-label image classification problem, more specifically, the subcellular localization of proteins from microscopy imagery. In this talk David Wagner will present his own approaches to the competition resulting in a top 15% placement, touch on the competitive edge of the winning solutions, and talk about lessons learned taking on a Kaggle challenge for the first time.

While this talk will not be a deep dive on deep learning or winning Kaggle competitions, the slides will be full of resources for those wanting to learn more on a number of subjects:
- Deep learning with Pytorch
- Convolution neural networks
- Approaches to transfer learning
- Visualization with Visdom and Facets Python libraries
- (Relatively) cheap compute using Google Cloud Preemptible VMs

David Wagner is a Machine Learning Engineer at Mycroft AI and before that spent three years as a Software Engineer at ZOLOZ (EyeVerify). He drank the Zen of Python Kool-Aid early in his career and is happiest solving problems inside of a Jupyter Notebook.

This meetup will be hosted at ZOLOZ.

ZOLOZ is located at 1740 Main St. Kansas City, MO on the bottom floor of the Barkley building with the rocket on top and the #KC Loves mural. Come early as you will need to park on the street in the Crossroads and may need to walk depending on how busy the area is. You must use the building's East entrance off of Main St. on the North side of 18th St.

Click here for event

0 Response to "June 26: Data Science KC - Competing in Kaggle's Protein Image Classification Challenge"

Post a Comment

Group Tools

Random Prize Winner
Use this tool to generate random numbers for prize drawings.

Computer Humor Slideshow
Just for fun!




Follow this twitter list of the twitter accounts for the user groups. Ask for your group to be added to this list: twitter list
Subscribe to the Kansas City User Group Newspaper at Paper.li

Blog Archive

Followers