Takayuki Akiyama

About

Takayuki Akiyama’s research centers on reinforcement learning, with a particular focus on sample efficiency and active learning for robotic control. His major contribution lies in bridging statistical active learning with value function approximation, demonstrating that the least-squares policy iteration (LSPI) framework can leverage active learning methods to design sampling policies that yield better control policies with fewer interactions. His most-cited work, “Efficient exploration through active learning for value function approximation in reinforcement learning” (2010, 20 citations), formalizes this approach and has become a reference point for researchers tackling exploration-exploitation trade-offs. Akiyama also applies deep reinforcement learning to practical robotics, as seen in his paper on experience filtering for robot navigation (2018, 7 citations), which addresses how to select and prioritize training data for more robust navigation policies. While his citation counts reflect a focused, early-stage career, his work is foundational for those developing sample-efficient RL algorithms—a critical challenge in deploying RL in real-world systems where data is costly. Akiyama’s research continues to influence how agents learn from limited experience.

Research Focus

Key Achievements

Top Papers

1. 1
   Efficient exploration through active learning for value function approximation in reinforcement learning

   20 citations · 2010
2. 2
   Experience Filtering for Robot Navigation using Deep Reinforcement Learning

   7 citations · 2018
3. 3
   Active policy iteration: efficient exploration through active learning for value function approximation in reinforcement learning

   2 citations · 2009

Key Collaborators

Contact & Links