Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Open Problem: Active Representation Learning

Nikola Milosevic, Gesine Müller, Jan Huisken, Nico Scherf

TL;DR

The need for a framework that derives exploration skills from representations that are in some sense actionable is explored, aiming to enhance the efficiency and effectiveness of data collection and model building in the natural sciences.

Abstract

In this work, we introduce the concept of Active Representation Learning, a novel class of problems that intertwines exploration and representation learning within partially observable environments. We extend ideas from Active Simultaneous Localization and Mapping (active SLAM), and translate them to scientific discovery problems, exemplified by adaptive microscopy. We explore the need for a framework that derives exploration skills from representations that are in some sense actionable, aiming to enhance the efficiency and effectiveness of data collection and model building in the natural sciences.

Open Problem: Active Representation Learning

TL;DR

The need for a framework that derives exploration skills from representations that are in some sense actionable is explored, aiming to enhance the efficiency and effectiveness of data collection and model building in the natural sciences.

Abstract

In this work, we introduce the concept of Active Representation Learning, a novel class of problems that intertwines exploration and representation learning within partially observable environments. We extend ideas from Active Simultaneous Localization and Mapping (active SLAM), and translate them to scientific discovery problems, exemplified by adaptive microscopy. We explore the need for a framework that derives exploration skills from representations that are in some sense actionable, aiming to enhance the efficiency and effectiveness of data collection and model building in the natural sciences.
Paper Structure (14 sections, 3 equations, 1 figure)

This paper contains 14 sections, 3 equations, 1 figure.

Table of Contents

  1. Introduction
  2. Related Work
  3. From Adaptive to Active Microscopy.
  4. Representations in Online Learning
  5. Representations in Deep Reinforcement Learning.
  6. Active Representation Learning
  7. Motivation
  8. Relation to Other Decision Frameworks
  9. From Active SLAM to Active Microscopy
  10. Challenges
  11. Opportunities
  12. Conclusion
  13. Extended Background
  14. Background on POMDPs and Active SLAM

Figures (1)

  • Figure 1: Active representation learning (ARL) is a set of problems that involves deriving Exploration Skills (ES), informed by Actionable Representations (AR), to learn a suitable exploration policy $\pi$, see Figure a. Two common examples are active SLAM (Figure b) and active microscopy (Figure c).