基本信息
浏览量:0
职业迁徙
个人简介
Research Interests
I am deeply interested in enabling decision-making agents to learn reusable knowledge. To that end, currently, I’m working on learning state abstractions for Reinforcement Learning and, more generally, MDPs, that allow agents to learn provably sound, abstract and simpler world models that the agent can use to generate plans for different tasks.
Furthermore, I’ve also worked at the intersection of Natural Language and RL, investigating how to communicate prior knowledge to RL agents through languages. From this endeavor, RLang, a formal language for RL, was born. This language is unambiguous and designed precisely for RL. RLang allows to communicate partial task-specific knowledge to RL agents in order to avoid tabula rasa learning. The RLang framework, also, opens up many exciting research questions in RL algorithm design, natural language understanding and symbol grounding.
I am deeply interested in enabling decision-making agents to learn reusable knowledge. To that end, currently, I’m working on learning state abstractions for Reinforcement Learning and, more generally, MDPs, that allow agents to learn provably sound, abstract and simpler world models that the agent can use to generate plans for different tasks.
Furthermore, I’ve also worked at the intersection of Natural Language and RL, investigating how to communicate prior knowledge to RL agents through languages. From this endeavor, RLang, a formal language for RL, was born. This language is unambiguous and designed precisely for RL. RLang allows to communicate partial task-specific knowledge to RL agents in order to avoid tabula rasa learning. The RLang framework, also, opens up many exciting research questions in RL algorithm design, natural language understanding and symbol grounding.
研究兴趣
论文共 3 篇作者统计合作学者相似作者
按年份排序按引用量排序主题筛选期刊级别筛选合作者筛选合作机构筛选
时间
引用量
主题
期刊级别
合作者
合作机构
作者统计
合作学者
合作机构
D-Core
- 合作者
- 学生
- 导师
数据免责声明
页面数据均来自互联网公开来源、合作出版商和通过AI技术自动分析结果,我们不对页面数据的有效性、准确性、正确性、可靠性、完整性和及时性做出任何承诺和保证。若有疑问,可以通过电子邮件方式联系我们:report@aminer.cn