Current browse context:
stat.ML
Change to browse by:
References & Citations
Statistics > Machine Learning
Title: Hierarchical Expert Networks for Meta-Learning
(Submitted on 31 Oct 2019 (v1), revised 4 Feb 2020 (this version, v6), latest version 9 Sep 2020 (v7))
Abstract: The goal of meta-learning is to train a model on a variety of learning tasks, such that it can adapt to new problems within only a few iterations. Here we propose a principled information-theoretic model that optimally partitions the underlying problem space such that the resulting partitions are processed by specialized expert decision-makers. To drive this specialization we impose the same kind of information processing constraints both on the partitioning and the expert decision-makers. We argue that this specialization leads to efficient adaptation to new tasks. To demonstrate the generality of our approach we evaluate on three meta-learning domains: image classification, regression, and reinforcement learning.
Submission history
From: Heinke Hihn [view email][v1] Thu, 31 Oct 2019 06:23:52 GMT (1225kb,D)
[v2] Thu, 7 Nov 2019 09:23:59 GMT (1225kb,D)
[v3] Wed, 13 Nov 2019 12:27:14 GMT (1189kb,D)
[v4] Thu, 14 Nov 2019 17:03:56 GMT (1189kb,D)
[v5] Wed, 4 Dec 2019 15:18:43 GMT (1189kb,D)
[v6] Tue, 4 Feb 2020 12:10:39 GMT (1223kb,D)
[v7] Wed, 9 Sep 2020 16:38:27 GMT (639kb,D)
Link back to: arXiv, form interface, contact.