The article presents a new framework for Interactive Machine Learning (IML), aiming to incorporate human expertise into machine learning algorithms. The framework is grounded in Bayesian Optimization (BO), thus called Interactive Bayesian Optimization (IBO), and makes use of a unique acquisition function, Preference Expected Improvement (PEI), to capture user preferences and enhance system efficiency. The IBO allows human-machine collaboration by providing an interface for users to shape the strategy by hand. This approach is viewed as an optimization task where humans and algorithms collaborate to refine the agent’s actions. The framework was applied in simulations and a real-world task using a Franka Panda robot to demonstrate human-robot collaboration.
Publication date: 25 Jan 2024
Project Page: Not Provided
Paper: https://arxiv.org/pdf/2401.12662
