On the Importance of Exploration for Real Life Learned Algorithms

Steffen Gracla, Carsten Bockelmann, Armin Dekorsy
4/21/2023
cs.LGcs.NIeess.SY

No official code found. Best implementation: Steffengra/on_exploration (0 stars, MIT, last commit 2025-02-07).

Abstract

The quality of data driven learning algorithms scales significantly with the quality of data available. One of the most straight-forward ways to generate good data is to sample or explore the data source intelligently. Smart sampling can reduce the cost of gaining samples, reduce computation cost in learning, and enable the learning algorithm to adapt to unforeseen events. In this paper, we teach three Deep Q-Networks (DQN) with different exploration strategies to solve a problem of puncturing ongoing transmissions for URLLC messages. We demonstrate the efficiency of two adaptive exploration candidates, variance-based and Maximum Entropy-based exploration, compared to the standard, simple epsilon-greedy exploration approach.

View on arXivView PDF

Code Implementations(1)

Steffengra/on_explorationCommunity100%
MIT
00PythonFeb 18, 20221 years agoMIT

Similar Papers

On the Convergence and Sample Complexity Analysis of Deep Q-Networks with ε-Greedy Exploration

85% match2026

DQN Performance with Epsilon Greedy Policies and Prioritized Experience Replay

85% match2025

β-DQN: Improving Deep Q-Learning By Evolving the Behavior

83% match1 repo2025

Value of Information-Enhanced Exploration in Bootstrapped DQN

83% match2025

Lifting the Veil: Unlocking the Power of Depth in Q-learning

81% match2023

Adaptive Q-Network: On-the-fly Target Selection for Deep Reinforcement Learning

81% match2024

Cite this paper

@article{gracla2023importance,
  title  = {On the Importance of Exploration for Real Life Learned Algorithms},
  author = {Steffen Gracla and Carsten Bockelmann and Armin Dekorsy},
  year   = {2023},
  eprint = {2304.10860},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2304.10860v1}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2304.10860.svg)](https://codeofpaper.com/paper/2304.10860)

Discussion