Seonglae ChoActive Learning focuses on improving the overall accuracy of the function model by selecting areas with high uncertainty estimated through Gaussian Process
Active Learning aims to model the entire function, while Bayesian Optimization aims to find the optimal point of a black-box function considering Acquisition Function while both use Gaussian Process as a surrogate model (with a prior over the space of objective functions) which serves as a surrogate for function approximation, uncertainty estimation, and uncertainty reduction.
It requests labeling for important learning instances among data that hasn't been labeled yet. Subsequently cycles through conducting learning using data rich in information and requesting additional data
- Choose and add the point with the highest uncertainty to the training set (by querying/labeling that point)
- Train on the new training set
- Go to #1 till convergence or budget elapsed
Active Learning Notion
Active Learning Methods
Exploring Bayesian Optimization
How to tune hyperparameters for your machine learning model using Bayesian optimization.
https://distill.pub/2020/bayesian-optimization
Active Learning - ① 액티브 러닝이란 무엇인가
전통적으로 기계학습(Passive Machine Learning)은 라벨링되지 않은 데이터에 대해 사람이 라벨을 부여하면 이를 기계가 학습하는 방식으로 이루어졌다. 이 방식에서는 학습 데이터 마련에 상당한 사람의 공수가 들어간다. 뿐만아니라 이미지넷 리더보드나 기계번역 모델을 보면 사람보다도 태스크를 잘 수행하는 모델이 등장하였는데, 이렇게 잠재적으로 뛰어난 기계를 두고 사람이 모든 라벨링을 진행하는 것은 조금 아깝다는 생각이 든다.
https://littlefoxdiary.tistory.com/52
