A Hybrid Multi-Task Learning Approach for Optimizing Deep Reinforcement Learning Agents

Driven by recent technological advancements within the field of artificial intelligence (AI), deep learning (DL) has been emerged as a promising representation technique across different machine (ML) classes, especially reinforcement (RL) arena. This new direction given rise to evolution domain named (DRL) that combines high representational capabilities DL with existing RL methods. Performance optimization achieved RL-based intelligent agents designed model-free-based approaches was majorly limited systems algorithms focused on single task. The aforementioned approach found be quite data inefficient, whenever DRL needed interact more complex, data-rich environments. is primarily due applicability many scenarios related tasks from same distribution. One possible mitigate this issue adopting method multi-task learning. objective research paper present hybrid learning-oriented for operating but semantically similar environments tasks. proposed framework will built multiple, individual actor-critic models functioning independent and transferring knowledge among themselves through global network optimize performance. empirical results obtained model OpenAI Gym based Atari 2600 video gaming environment demonstrates enhances performance agent relatively in range 15% 20% margin.