Towards continual learning in medical imaging

File Description SizeFormat 
1811.02496v1.pdf1.43 MBAdobe PDFView/Open
Title: Towards continual learning in medical imaging
Authors: Baweja, C
Glocker, B
Kamnitsas, K
Item Type: Working Paper
Abstract: This work investigates continual learning of two segmentation tasks in brain MRI with neural networks. To explore in this context the capabilities of current methods for countering catastrophic forgetting of the first task when a new one is learned, we investigate elastic weight consolidation, a recently proposed method based on Fisher information, originally evaluated on reinforcement learning of Atari games. We use it to sequentially learn segmentation of normal brain structures and then segmentation of white matter lesions. Our findings show this recent method reduces catastrophic forgetting, while large room for improvement exists in these challenging settings for continual learning.
URI: http://hdl.handle.net/10044/1/71278
Keywords: cs.CV
cs.CV
cs.LG
Notes: Accepted in Medical Imaging meets NIPS Workshop, NIPS 2018
Appears in Collections:Faculty of Engineering
Computing



Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.

Creative Commonsx