<feed xmlns="http://www.w3.org/2005/Atom"> <id>https://danielredglasses.github.io/</id><title>Daniel Park</title><subtitle>A personal blog documenting my journey in Reinforcement Learning and Competitive Programming.</subtitle> <updated>2026-06-24T14:51:08+09:00</updated> <author> <name>Daniel Park</name> <uri>https://danielredglasses.github.io/</uri> </author><link rel="self" type="application/atom+xml" href="https://danielredglasses.github.io/feed.xml"/><link rel="alternate" type="text/html" hreflang="en" href="https://danielredglasses.github.io/"/> <generator uri="https://jekyllrb.com/" version="4.4.1">Jekyll</generator> <rights> © 2026 Daniel Park </rights> <icon>/assets/img/favicons/favicon.ico</icon> <logo>/assets/img/favicons/favicon-96x96.png</logo> <entry><title>DUSDi (Disentangled Unsupervised Skill Discovery)</title><link href="https://danielredglasses.github.io/en/posts/DUSDi/" rel="alternate" type="text/html" title="DUSDi (Disentangled Unsupervised Skill Discovery)" /><published>2026-06-17T16:15:35+09:00</published> <updated>2026-06-17T16:15:35+09:00</updated> <id>https://danielredglasses.github.io/en/posts/DUSDi/</id> <content type="text/html" src="https://danielredglasses.github.io/en/posts/DUSDi/" /> <author> <name>Daniel Park</name> </author> <category term="Reinforcement Learning" /> <summary>Disentangled Unsupervised Skill Discovery (DUSDi) is a method for learning disentangled skills in factored state spaces. What does it mean by the skills are disentangled? The paper gives an example of driving a car to compare between prior work and DUSDi. If a single skill variable simultaneously changes the speed, steering, and headlights, it will be extremely challenging to learn how to tu...</summary> </entry> <entry><title>DUSDi (Disentangled Unsupervised Skill Discovery)</title><link href="https://danielredglasses.github.io/ko/posts/DUSDi/" rel="alternate" type="text/html" title="DUSDi (Disentangled Unsupervised Skill Discovery)" /><published>2026-06-17T16:15:35+09:00</published> <updated>2026-06-17T16:15:35+09:00</updated> <id>https://danielredglasses.github.io/ko/posts/DUSDi/</id> <content type="text/html" src="https://danielredglasses.github.io/ko/posts/DUSDi/" /> <author> <name>Daniel Park</name> </author> <category term="Reinforcement Learning" /> <summary>DUSDi(Disentangled Unsupervised Skill Discovery)는 요인화된 상태 공간(factored state space)에서 분리된(disentangled) 스킬을 학습하는 방법입니다. 스킬이 분리되어 있다는 것은 무슨 의미일까요? 논문에서는 자동차 운전을 예시로 들어 기존 연구와 DUSDi를 비교합니다. 하나의 스킬 변수가 속도, 조향, 헤드라이트를 동시에 바꾼다면, 차의 속도와 방향을 유지하면서 헤드라이트만 켜고 끄는 것은 여러 스킬 변수를 복잡하게 조합해야 하므로 매우 어려운 일이 됩니다. 하지만 스킬 변수 $z_1$이 속도만, $z_2$가 조향만, $z_3$가 헤드라이트만 제어한다면, 에이전트는 $z_1$과 $z_2$를 고정한 채 $z_3$만 조절하여 위 명령을 ...</summary> </entry> </feed>
