Active few shot learning: The future of training Machine Learning models
Performance
• March 2021
Active few shot learning: The future of training Machine Learning models
About
Active few shot learning: The future of training Machine Learning models
About
Few-shot learning addresses the problem of learning new, unseen concepts quickly with limited number of annotated training samples. Active learning is based on the idea that smart sampling of data leads to faster training and more accurate models.
Today, unlabeled data is present in abundance while obtaining labeled data is costly and tedious. By combining the power of giant pre-trained models, and the capabilities of active learning to aid the data annotation process, we can build high performing domain specific models by using Active few shot learning.
Language
English
Level
Intermediate
Length
29 minutes
Type
online conference
About the speaker
About the speaker
Surabhi Bhargava
Machine Learning Scientist •
Adobe
I am a Machine Learning Scientist at Adobe focused on building product intelligence for improved user experience. I am passionate about working on Computer Vision and Natural Language Processing applications and have published work in these areas throughout my undergraduate and graduate studies. I have worked extensively with computer vision techniques for natural images and documents as well as multimodal applications. I want to share my knowledge and experience regarding such techniques, with AI and ML becoming so popular with others in my community to inspire and build solutions together.
Details
Language
English
Level
Intermediate
Length
29 minutes
Type
online conference