首页|New Findings in Robotics and Automation Described from Budapest University of Te chnology and Economics (Adaptive Curriculum Learning With Successor Features for Imbalanced Compositional Reward Functions)
New Findings in Robotics and Automation Described from Budapest University of Te chnology and Economics (Adaptive Curriculum Learning With Successor Features for Imbalanced Compositional Reward Functions)
扫码查看
点击上方二维码区域，可以放大扫码查看
原文链接
NETL
NSTL
By a News Reporter-Staff News Editor at Robotics & Machine Learning Daily News Daily News – New research on Robotics - Robotics an d Automation is the subject of a report. According to news reporting out of Buda pest, Hungary, by NewsRx editors, research stated, “This work addresses the chal lenge of reinforcement learning with reward functions that feature highly imbala nced components in terms of importance and scale. Reinforcement learning algorit hms generally struggle to handle such imbalanced reward functions effectively.” Financial support for this research came from Ministry of Culture and Innovation of Hungary from the National Research, Development, and Innovation Fund. Our news journalists obtained a quote from the research from the Budapest Univer sity of Technology and Economics, “Consequently, they often converge to suboptim al policies that favor only the dominant reward component. For example, agents m ight adopt passive strategies, avoiding any action to evade potentially unsafe o utcomes entirely. To mitigate the adverse effects of imbalanced reward functions , we introduce a curriculum learning approach based on the successor features re presentation.” According to the news editors, the research concluded: “This novel approach enab les our learning system to acquire policies that take into account all reward co mponents, allowing for a more balanced and versatile decision-making process.”
BudapestHungaryEuropeRobotics and AutomationRoboticsEmerging TechnologiesMachine LearningReinforcement Lea rningBudapest University of Technology and Economics
NETL
NSTL
