HOME

Combinatorial Quantum Materials Research Lab

Department of Physics, Pukyong National University

How to teach a baby what a pig is?

Even if the image is distorted, as long as the key feature — the pig’s snout — remains,
and we show the baby many such images while saying “This is a pig,”
the baby will eventually learn that the snout is the most important feature for recognizing a pig.

This is the core idea behind data augmentation in machine learning.

By distorting irrelevant parts while preserving essential physical features,
we help the model focus on what truly matters —
a strategy we call physics-guided data augmentation.

NEWS from CQMLab

우리 연구실 석사과정 이동익 학생이 제1저자로 참여한 물리학 기반
데이터 증강 전략을 활용한 머신러닝 기반 초전도체 스펙트럼 분석 연구가
Materials Today Physics 지에 게재되었습니다!

Our first machine learning paper is out!

It's about building a machine learning model for very rapid analysis of point-contact spectra. The performance is significantly improved through physics-based data augmentation!

It’s now published in Materials Today Physics! I’m so happy!

📢 ChatGPT를 어떻게 잘 활용할 것인가?

우리 연구실 석사과정 이유경 학생이 제1저자로 참여한 ChatGPT 프롬프트 엔지니어링(prompt engineering) 관련 연구가
Scientific Reports 지에 게재되었습니다!

“… We also emphasize the importance of having a certain level of proper knowledge for designing the prompts and performing minimal cross-checking …”
— from the paper

It’s now published in Scientific Reports! 🎉

우리 연구실의 오주현 박사님(제1저자), 남기득, 이동익 학생이 참여한
스퍼터링 방식을 통한 최초의 antiperovskite 박막 제작과 조합방식을 통한 구조적 특성 연구가 Appl. Phys. Lett. 지에 게재 확정되었습니다! :)

(Our work on the first fabrication of an antipervskite oxide using sputtering, followed by a combinatorial study on the stoichiometry effect, has been accepted for publication in Applied Physics Letters! Congratulations!)

NEW Article on KPS Webzine
(물리학과 첨단기술)