Unifying Approaches in Active Learning and Active Sampling via Fisher Information and Information-Theoretic Quantities | PDF

N/A

Published in Transactions on Machine Learning Research (11/2022) Unifying Approaches in Active Learning and Active Sampling via Fisher Information and Information-Theoretic Quantities Andreas Kirsch andreas.kirsch@cs.ox.ac.uk Yarin Gal yarin.gal@cs.ox.ac.uk…

related reading

• An Introduction to Fisher Information – Awni Hannun – Writing About Machine Learningawni.github.io
• Bayesian Neural Networkscs.toronto.edu
• Learning with not Enough Data Part 2: Active Learninglilianweng.github.io
• The Little Book of Deep Learningfleuret.org
• A Retrospective on Active Inferenceberen.io
• Gregory Gundersengregorygundersen.com
• Competing with sampling — Alignment Research Centeralignment.org
• David Duvenaudcs.toronto.edu
• Free energy principle - Wikipediaen.wikipedia.org
• Contra Dwarkesh on RL sample-efficiency via information theorynewsletter.danielpaleka.com
• [2109.03853] A Bayesian Framework for Information-Theoretic Probingarxiv.org
• [1809.09087] Implicit Maximum Likelihood Estimationarxiv.org