Welcome to the last Active Learning newsletter of 2020! Before we get into the hot-off-the-arXiv-press research, I just want to let you know about a webinar that I will be co-hosting with an NLP-focused AI startup Kairntech on December 17, 17:30 Paris-time (also known as 5:30 pm CET):
What does “active learning” mean in AI? Save time and money on data annotation. The Kairntech…
If you are new to active machine learning, then (a) it is unlikely that you made it this far down, (b) you might want to check out my introductory blog posts on this subject:
Or, for the more audio-visually minded, here is my recent beginner-friendly conference talk:
Without further ado, let’s get to the advanced business.
Healthcare is one of the domains where the costs associated with manual data annotation tend to be particularly high, so it is always nice to see an active learning benchmark for a medical application. The authors of A Transfer Learning Based Active Learning Framework for Brain Tumor Classification used active learning to develop a model than performed slightly better than the fully supervised baseline while reducing the number of the labelled samples used for training by an impressive forty percent!
#ai #deep-learning #data-science
